Skip to main content

Full text of "Course in general linguistics"

See other formats









Edited  by  CHARLES  BALLY  and 

In  collaboration  with 

Translated  from  the  French  by  WADE  BASKIN 


New  York 

COPYRIGHT,    1959,   BY 


15   EAST   40th   street,    new   YORK   CITY 

Printed  in  the  United  States  of  America 







v^s-  l=r^ 







Translator's  Introduction xi 

Preface  to  the  First  Edition xiii 



I.  A  Glance  at  the  History  of  Linguistics [v) 

II.  Subject  Matter  and  Scope  of  Linguistics ;  Its  Relations 

with  Other  Sciences 6 

III.  Object  of  Linguistics 

1.  Definition  of  Language 7 

2.  Place  of  Language  in  the  Facts  of  Speech  ...  11 

3.  Place  of  Language  in  Human  Facts;  Semiology  .  \i5{ 

IV.  Linguistics  of  Language  and  Linguistics  of  Speaking  .  17 
V.  Internal  and  External  Elements  of  Language  .      .      .  \.2Q 

VI.  Graphic  Representation  of  Language  '"'^^ 

1.  Need  for  Studying  the  Subject 23 

2.  Influence  of  Writing;  Reason  for  Its  Ascendancy 
over  the  Spoken  Form 23 

3.  Systems  of  Writing 25 

4.  Reasons  for  the  DiscrepfUncy  between  Writing  and 
Pronunciation 27 

5.  Results  of  the  Discrepancy 29 

VII.  Phonology 

1.  Definition 32 

2.  Phonological  Writing 33 

3.  Validity  of  Evidence  Furnished  by  Writing     .      .  34 






Chapter  Page 

I.  Phonological  Species 

1.  Definition  of  the  Phoneme 38 

2.  The  Vocal  Apparatus  and  its  Functioning  ...  41 

3.  Classification  of  Sounds  According  to  Their  Oral 
Articulation 44 

11.  Phonemes  in  the  Spoken  Chain 

L  Need  for  Studying  Sounds  in  the  Spoken  Chain  .  49 

2.  Implosion  and  Explosion 51 

3.  Different   Combinations   of   Explosions   and   Im- 
plosions in  the  Chain 54 

4.  Syllabic  Boundary  and  Vocalic  Peak  ....  57 

5.  Criticism  of  Theories  of  Syllabication  ....  58 

6.  Length  of  Implosion  and  Explosion 60 

7.  Phonemes  of  Aperture  4;  Diphthongs;  Questions 

about  Transcription 60 

Editor's  Note 62 



I.  Nature  of  the  Linguistic  Sign  r"^ 

1.  Sign,  Signified,  Signifier 1  65j 

2.  Principle  I:  The  Arbitrary  Nature  of  the  Sign     .  \^ 

3.  Principle  II:  The  Linear  Nature  of  the  Signifier  .  70 
11.  Immutability  and  Mutability  of  the  Sign  ^ — ^ 

1.  Immutability \  71 

2.  Mutability [JUj 

III.  Static  and  Evolutionary  Linguistics 

1.  Inner  Duality   of   All   Sciences   Concerned   with 
Values 79 

2.  Inner  Duality  and  the  History  of  Linguistics  .     .  81 

3.  Inner  Duality  Illustrated  by  Examples  ....  83 


Chapter  Page 

4.  The  Difference  between  the  Two  Classes  Illustrated 

by  Comparisons 87 

5.  The  Two  Linguistics  Contrasted  According  to  Their 
Methods  and  Principles 90 

6.  Synchronic  Law  and  Diachronic  Law  .      .      .      .  '  91^ 

7.  Is  There  a  Panchronic  Viewpoint? "95 

8.  Consequences  of  the  Confusing  of  Synchrony  and 
Diachrony 96/ 

9.  Conclusions 98 



I.  Generalities 101 

11.  The  Concrete  Entities  of  Language 

1.  Definition  of  Entity  and  Unit 102 

2.  Method  of  Delimitation 104 

3.  Practical  Difficulties  of  Delimitation   ....  105 

4.  Conclusion 106 

III.  Identities,  Realities,  Values 107  , 

IV.  Linguistic  Value 

1.  Language   as   Organized   Thought   Coupled   with  -  ^ 
Sound rill 

2.  Linguistic  Value  from  a  Conceptual  Viewpoint    .  iJu4i 

3.  Linguistic  Value  from  a  Material  Viewpoint  .      .  HZ 

4.  The  Sign  Considered  in  Its  Totality     ....  '>120} 
V.  Syntagmatic  and  Associative  Relations  "" 

1.  Definitions 122 

2.  Syntagmatic  Relations 124 

3.  Associative  Relations 125 

VI.  Mechanism  of  Language 

1.  Syntagmatic  Solidarities 127 

2.  Simultaneous  Functioning  of  the  Two  Types  of 
Groupings 1^8, 

3.  Absolute  and  Relative  Arbitrariness     ....  \131/ 


Chapter  Page 

VII.  Grammar  and  Its  Subdivisions 

1.  Definitions:  Traditional  Divisions 134 

2.  Rational  Divisions 136 

VIII.  Role  of  Abstract  Entities  in  Grammar 137 



I.  Generalities 140 

II.  Phonetic  Changes 

1.  Their  Absolute  Regularity 143 

2.  Conditioned  Phonetic  Changes 144 

3.  Points    on    Method 145 

4.  Causes  of  Phonetic  Changes 147 

5.  The  Effect  of  Phonetic  Changes  Is  Unlimited  .      .  151 

III.  Grammatical  Consequences  of  Phonetic  Evolution  __ 

1.  Breaking  of  the  Grammatical  Bond :  ^^J 

2.  Effacement  of  the  Structure  of  Words  ....  flSi- 

3.  There  are  No  Phonetic  Doublets 155 

4.  Alternation 157 

5.  Laws  of  Alternation 158 

6.  Alternation  and  Grammatical  Bond 160 

IV.  Analogy 

1.  Definition  and  Examples 161 

2.  Analogical  Phenomena  Are  Not  Changes  .      .      .  162 

3.  Analogy  as  a  Creative  Force  in  Language  .     .      .  165 
V.  Analogy  and  Evolution 

1.  How  an  Analogical  Innovation  Enters  Language  .  ■168 

2.  Analogical  Innovations  as  Symptoms  of  Changes 

in  Interpretation 169 

3.  Analogy  as  a  Renovating  and  Conservative  Force  171 
VI.  Folk  Etymology 173 

VII.  Agglutination 

1.  Definition 176 

2.  Agglutination  and  Analogy 177 


Chapter  Page 

VIII.  Diachronic  Units,  Identities,  and  Realities   .      .      .  179 
Appendices  to  Parts  Three  and  Four 

1.  Subjective  and  Objective  Analysis 173 

2.  Subjective  Analysis  and  the  Defining  of  Subunits  185 

3.  Etymology 189 



I.  Concerning  the  Diversity  of  Languages   ....     191 
II.  Complications  of  Geographical  Diversity 

1.  Coexistence  of  Several  Languages  at  the   Same 
Point 193 

2.  Literary  Language  and  Local  Idiom   ....     195 

III.  Causes  of  Geographical  Diversity 

1.  Time,  the  Basic  Cause 197 

2.  Effect  of  Time  on  Continuous  Territory  ,      .      .  199 

3.  Dialects  Have  No  Natural  Boundaries     .      .      .  201 

4.  Languages  Have  No  Natural  Boundaries  .      .      .  203 

IV.  Spread  of  Linguistic  Waves 

1.  Intercourse  and  Provincialism 205 

2.  The  Two  Forces  Reduced  to  One 207 

3.  Linguistic  Differentiation  on  Separate  Territories    208 



I.  The  Two  Perspectives  of  Diachronic  Linguistics  .      .     212 
11.  The  Oldest  Language  and  the  Prototype  ....     215 

III.  Reconstructions 

1.  Their  Nature  and  Aim 218 

2.  Relative  Accuracy  of  Reconstructions  ....     220 

IV.  The  Contribution  of  Language  to  Anthropology  and 

1.  Language  and  Race 222 

2.  Ethnic   Unity 223 


Chapter  Page 

3.  Linguistic  Paleontology 224 

4.  Linguistic  Type  and  Mind  of  the  Social  Group  .     227 
V.  Language  Families  and  Linguistic  Types  ....     228 

Index 233 


Few  other  figures  in  the  history  of  the  science  of  language  have 
commanded  such  lasting  respect  and  inspired  such  varied  accom- 
phshments  as  Ferdinand  de  Saussure.  Leonard  Bloomfield  justly 
credited  the  eminent  Swiss  professor  with  providing  "a  theoretic 
foundation  to  the  newer  trend  in  Unguistics  study,"  and  European 
scholars  have  seldom  failed  to  consider  his  views  when  deaUng 
with  any  theoretical  problem.  But  the  full  implications  of  his 
teachings,  for  both  static  and  evolutionary  studies,  have  still  to 
be  elaborated. 

Saussure  succeeded  in  impressing  his  individual  stamp  on 
almost  everything  within  his  reach.  At  the  age  of  twenty,  while 
still  a  student  at  Leipzig,  he  published  his  monumental  treatise 
on  the  Proto-Indo-European  vocalic  system.  This  treatise,  though 
based  on  theories  and  facts  that  were  common  property  in  his 
day,  is  still  recognized  as  the  most  inspired  and  exhaustive  treat- 
ment of  the  Proto-Indo-European  vocalism.  He  studied  under 
the  neogrammarians  Osthoff  and  Leskien,  yet  refuted  their  atom- 
istic approach  to  linguistics  in  his  attempt  to  frame  a  coherent 
science  of  linguistics.  Despite  the  paucity  of  his  publications  (some 
600  pages  during  his  lifetime),  Saussure's  influence  has  been  far- 
reaching.  At  Paris,  where  he  taught  Sanskrit  for  ten  years  (1881- 
1891)  and  served  as  secretary  of  the  Linguistic  Society  of  Paris, 
his  influence  on  the  development  of  hnguistics  was  decisive.  His 
first-hand  studies  of  Phrygian  inscriptions  and  Lithuanian  dialects 
may  have  been  responsible  for  some  of  the  quahties  that  subse- 
quently endeared  him  to  his  students  at  the  University  of  Geneva 
(1906-1911).  His  unique  insight  into  the  phenomenon  of  language 
brought  to  fruition  the  best  of  contemporary  thinking  and  long 
years  of  patient  investigation  and  penetrating  thought. 

The  dominant  philosophical  system  of  each  age  makes  its 
imprint  on  each  step  in  the  evolution  of  linguistic  science.  The 
nineteenth  century  had  a  fragmentary  approach  to  reality  which 
prevented  scholars  from  getting  beyond  the  immediate  facts  in 


matters  of  speech.  To  those  investigators,  language  was  simply 
an  inventory  or  mechanical  sum  of  the  units  used  in  speaking. 
Piecemeal  studies  precluded  the  development  of  an  insight  into 
the  structure  (Gestalteinheit,  pattern,  or  whole)  into  which  the 
fragmentary  facts  fit.  The  atomistic  conception  of  speech,  reflected 
in  the  historical  studies  of  the  comparative  philologists,  had  to 
give  way  to  the  functional  and  structural  conception  of  language. 
Saussure  was  among  the  first  to  see  that  language  is  a  self-con- 
tained system  whose  interdependent  parts  function  and  acquire 
value  through  their  relationship  to  the  whole. 

By  focusing  attention  on  the  distinctly  human  side  of  speech, 
i.e.  the  system  'of  language,  Saussure  gave  unity  and  direction  to 
his  science.  Until  the  publication  of  his  work  (later  translated 
into  German  and  Spanish),  only  those  who  enjoyed  the  privilege 
of  close  association  with  Saussure  had  access  to  his  theories.  By 
making  available  an  English  translation  of  his  Course,  I  hope  to 
contribute  toward  the  reaUzation  of  his  goal:  the  study  of  language 
in  and  for  itself. 

To  all  those  who  have  given  generously  of  their  time  and  talents 
in  the  preparation  of  this  translation,  I  offer  heartfelt  thanks:  to 
Gerald  Dykstra,  Daniel  Girard,  Lennox  Grey,  Aileen  Kitchin, 
and  Andr^  Martinet  of  Columbia  University ;  to  Charles  Bazell  of 
Istanbul  University;  to  Henri  Frei,  Robert  Godel,  and  Edmond 
Sollberger  of  the  University  of  Geneva ;  to  Dwight  Bolinger  of  the 
University  of  Southern  California;  to  Rulon  Wells  of  Yale  Uni- 
versity; and  to  my  good  friends  Kenneth  Jimenez,  Paul  Swart, 
and  Hugh  Whittemore.  For  the  shortcomings  of  the  translation, 
I  alone  am  responsible. 

Wade  Baskin 


We  have  often  heard  Ferdinand  de  Saussure  lament  the  dearth  of 
principles  and  methods  that  marked  linguistics  during  his  develop- 
mental period.  Throughout  his  lifetime,  he  stubbornly  continued 
to  search  out  the  laws  that  would  give  direction  to  his  thought 
amid  the  chaos.  Not  until  1906,  when  he  took  the  place  of  Joseph 
Wertheimer  at  the  University  of  Geneva,  was  he  able  to  make 
known  the  ideas  that  he  had  nurtured  through  so  many  years. 
Although  he  taught  three  courses  in  general  hnguistics — in  1906- 
1907,  1908-1909,  and  1910-1911— his  schedule  forced  him  to  de- 
vote half  of  each  course  to  the  history  and  description  of  the  Indo- 
European  languages,  with  the  result  that  the  basic  part  of  his 
subject  received  considerably  less  attention  than  it  merited. 

All  those  who  had  the  privilege  of  participating  in  his  richly 
rewarding  instruction  regretted  that  no  book  had  resulted  from  it. 
After  his  death,  we  hoped  to  find  in  his  manuscripts,  obligingly 
made  available  to  us  by  Mme.  de  Saussure,  a  faithful  or  at  least 
an  adequate  outline  of  his  inspiring  lectures.  At  first  we  thought 
that  we  might  simply  collate  F.  de  Saussure's  personal  notes  and 
the  notes  of  his  students.  We  were  grossly  misled.  We  found 
nothing — or  almost  nothing — that  resembled  his  students'  note- 
books. As  soon  as  they  had  served  their  purpose,  F.  de  Saussure 
destroyed  the  rough  drafts  of  the  outlines  used  for  his  lectures.  In 
the  drawers  of  his  secretary  we  found  only  older  outlines  which, 
although  certainly  not  worthless,  could  not  be  integrated  into  the 
material  of  the  three  courses. 

Our  discovery  was  all  the  more  disappointing  since  professorial 
duties  had  made  it  impossible  for  us  to  attend  F.  de  Saussure's 
last  lectures — and  these  mark  just  as  brilliant  a  step  in  his  career 
as  the  much  earlier  one  that  had  witnessed  the  appearance  of  his 
treatise  on  the  vocalic  system  of  Proto-Indo-European. 

We  had  to  fall  back  on  the  notes  collected  by  students  during 
the  course  of  his  three  series  of  lectures.  Very  complete  notebooks 
were  placed  at  our  disposal:  for  the  first  two  courses,  by  Messrs. 


Louis  Caille,  Leopold  Gautier,  Paul  Regard,  and  Albert  Riedlinger; 
for  the  third — the  most  important — by  Mme.  Albert  Sechehaye 
and  by  Messrs.  George  D^gallier  and  Francis  Joseph.  We  are  in- 
debted to  M.  Louis  Brtitsch  for  notes  on  one  special  point.  All  these 
contributors  deserve  our  sincere  thanks.  We  also  wish  to  express 
our  profound  gratitude  to  M.  Jules  Ronjat,  the  eminent  Romance 
scholar,  who  was  kind  enough  to  review  the  manuscript  before 
printing,  and  whose  suggestions  were  invaluable. 

What  were  we  to  do  with  our  materials?  First,  the  task  of 
criticism.  For  each  course  and  for  each  detail  of  the  course,  we 
had  to  compare  all  versions  and  reconstruct  F.  de  Saussure's 
thought  from  faint,  sometimes  conflicting,  hints.  For  the  first  two 
courses  we  were  able  to  enlist  the  services  of  M.  RiedUnger,  one 
of  the  students  who  have  followed  the  thought  of  the  master 
with  the  greatest  interest;  his  work  was  most  valuable.  For  the 
third  course  one  of  us,  A.  Sechehaye,  performed  the  same  detailed 
task  of  collating  and  synthesizing  the  material. 

But  after  that?  Oral  delivery,  which  is  often  contradictory  in 
form  to  written  exposition,  posed  the  greatest  difficulties.  Besides, 
F.  de  Saussure  was  one  of  those  men  who  never  stand  still;  his 
thought  evolved  in  all  directions  without  ever  contradicting  itself 
as  a  result.  To  publish  everything  in  the  original  form  was  impos- 
sible; the  repetitions — inevitable  in  free  oral  presentation — over- 
lappings,  and  variant  formulations  would  lend  a  motley  appear- 
ance to  such  a  publication.  To  limit  the  book  to  a  single  course — 
and  which  one? — was  to  deprive  the  reader  of  the  rich  and  varied 
content  of  the  other  two  courses;  by  itself  the  third,  the  most 
definitive  of  the  three  courses,  would  not  give  a  complete  account- 
ing of  the  theories  and  methods  of  F.  de  Saussure. 

One  suggestion  was  that  we  publish  certain  particularly  original 
passages  without  change.  This  idea  was  appealing  at  first,  but 
soon  it  became  obvious  that  we  would  be  distorting  the  thought 
of  our  master  if  we  presented  but  fragments  of  a  plan  whose  value 
stands  out  only  in  its  totality. 

We  reached  a  bolder  but  also,  we  think,  a  more  rational  solution : 
to  attempt  a  reconstruction,  a  synthesis,  by  using  the  third  course 
as  a  starting  point  and  by  using  all  other  materials  at  our  disposal, 
including  the  personal  notes  of  F.  de  Saussure,  as  supplementary 


sources.  The  problem  of  re-creating  F.  de  Saussure's  thought  was 
all  the  more  difficult  because  the  re-creation  had  to  be  wholly 
objective.  At  each  point  we  had  to  get  to  the  crux  of  each  particu- 
lar thought  by  trying  to  see  its  definitive  form  in  the  light  of  the 
whole  system.  We  had  first  to  weed  out  variations  and  irregu- 
larities characteristic  of  oral  delivery,  then  to  fit  the  thought  into 
its  natural  framework  and  present  each  part  of  it  in  the  order 
intended  by  the  author  even  when  his  intention,  not  always 
apparent,  had  to  be  surmised. 

From  this  work  of  assimilation  and  reconstruction  was  born  the 
book  that  we  offer,  not  without  apprehension,  to  the  enlightened 
public  and  to  all  friends  of  linguistics. 

Our  aim  was  to  draw  together  an  organic  whole  by  omitting 
nothing  that  might  contribute  to  the  overall  impression.  But  for 
that  very  reason,  we  shall  probably  be  criticized  on  two  counts. 

First,  critics  will  say  that  this  "whole"  is  incomplete.  In  his 
teaching  the  master  never  pretended  to  examine  all  parts  of  lin- 
guistics or  to  devote  the  same  attention  to  each  of  those  examined ; 
materially,  he  could  not.  Besides,  his  main  concern  was  not  that. 
Guided  by  some  fundamental  and  personal  principles  which  are 
found  everywhere  in  his  work — and  which  form  the  woof  of  this 
fabric  which  is  as  solid  as  it  is  varied — he  tried  to  penetrate ;  only 
where  these  principles  find  particularly  striking  applications  or 
where  they  apparently  conflict  with  some  theory  did  he  try  to 

That  is  why  certain  disciplines,  such  as  semantics,  are  hardly 
touched  upon.  We  do  not  feel  that  these  lacunae  detract  from  the 
overall  architecture.  The  absence  of  a  "hnguistics  of  speaking"  is 
regrettable.  This  study,  which  had  been  promised  to  the  students 
of  the  third  course,  would  doubtlessly  have  had  a  place  of  honor; 
why  his  promise  could  not  be  kept  is  too  well  known.  All  we  could 
do  was  to  collect  the  fleeting  impressions  from  the  rough  outlines 
of  this  project  and  put  them  into  their  natural  place. 

Conversely,  critics  may  say  that  we  have  reproduced  facts 
bearing  on  points  developed  by  F.  de  Saussure's  predecessors.  Not 
everything  in  such  an  extensive  treatise  can  be  new.  But  if  known 
principles  are  necessary  for  the  understanding  of  a  whole,  shall  we 
be  condemned  for  not  having  omitted  them?  The  chapter  on 


phonetic  changes,  for  example,  includes  things  that  have  been 
said  before,  and  perhaps  more  definitively;  but,  aside  from  the 
fact  that  this  part  contains  many  valuable  and  original  details, 
even  a  superficial  reading  will  show  to  what  extent  its  omission 
would  detract  from  an  understanding  of  the  principles  upon  which 
F.  de  Saussure  erects  his  system  of  static  hnguistics. 

We  are  aware  of  our  responsibility  to  our  critics.  We  are  also 
aware  of  our  responsibility  to  the  author,  who  probably  would  not 
have  authorized  the  publication  of  these  pages. 

This  responsibility  we  accept  wholly,  and  we  would  willingly 
bear  it  alone.  Will  the  critics  be  able  to  distinguish  between  the 
teacher  and  his  interpreters?  We  would  be  grateful  to  them  if  they 
would  direct  toward  us  the  blows  which  it  would  be  unjust  to  heap 
upon  one  whose  memory  is  dear  to  us. 

Geneva,  July  1915.  Charles  Bally,  Albert  Sechehaye 


The  second  edition  is  essentially  the  same  as  the  first.  The 
editors  have  made  some  slight  changes  designed  to  facilitate 
reading  and  clarify  certain  points.  Ch.  B.     Alb.  S. 


With  the  exception  of  a  few  minute  corrections,  this  edition  is 
the  same  as  the  preceding.  Ch.  B.    Alb.  S. 


Chapter  I 

The  science  that  has  been  developed  around  the  facts  of  language 
passed  through  three  stages  before  finding  its  true  and  unique 

First  something  called  "grammar"  was  studied.  This  study,  in- 
itiated by  the  Greeks  and  continued  mainly  by  the  French,  was 
based  on  logic.  It  lacked  a  scientific  approach  and  was  detached 
from  language  itself.  Its  only  aim  was  to  give  rules  for  distinguish- 
ing between  correct  and  incorrect  forms;  it  was  a  normative  dis- 
cipHne,  far  removed  from  actual  observation,  and  its  scope  was 

Next  appeared  philology.  A  "philological"  school  had  existed 
much  earlier  in  Alexandria,  but  this  name  is  more  often  applied 
to  the  scientific  movement  which  was  started  by  Friedrich  August 
Wolf  in  1777  and  which  continues  to  this  day.  Language  is  not  its 
sole  object.  The  early  philologists  sought  especially  to  correct, 
interpret  and  comment  upon  written  texts.  Their  studies  also  led 
to  an  interest  in  literary  history,  customs,  institutions,  etc.^  They 
apphed  the  methods  of  criticism  for  their  own  purposes.  When 
they  dealt  with  linguistic  questions,  it  was  for  the  express  purpose 
of  comparing  texts  of  different  periods,  determining  the  language 
peculiar  to  eacK^auihor,  or  deciphering  and  explaining  inscriptions 
made  in  an  archaic  or  obscure  language.  Doubtless  these  investi- 
gations broke  the  ground  for  historical  linguistics.  Rit^chl'^studies 
of  Plautus  are  actually  linguistic.\put  philological  criticism  is  still 
deficient  on  one  point:  it  follows  the  written  language  too  slavishly 

1  At  the  risk  of  offending  some  readers,  certain  stylistic  characteristics  of 
the  original  French  are  retained.  [Tr.]  (The  bracketed  abbreviations  S.,  Ed. 
and  Tr.  indicate  whether  footnotes  are  to  be  attributed  to  Saussure,  to  the 
editors  of  the  Cours  de  linguistique  generale,  or  to  the  translator.) 



and  neglects  the  living  language.  Moreover,  it  is  concerned  with 
little  except  Greek  and  Latin  antiquity. 

The  third  stage  began  when  scholars  discovered  that  languages 
can  be  compared  with  one  another.  This  discovery  was  the  origin 
of  "comparative  philology."  In  1816,  in  a  work  entitled  tJher  das 
Conjugationssijstem  der  Sanskritsprache,  Franz  Bopp  compared 
Sanskrit  with  German,  Greek,  Latin,  etc.  Bopp  was  not  the  first 
to  record  their  similarities  and  state  that  all  these  languages  belong 
to  a  single  family.  That  had  been  done  before  him,  notably  by  the 
English  orientalist  W.  Jones  (died  in  1794) ;  but  Jones'  few  isolated 
statements  do  not  prove  that  the  significance  and  importance  of 
comparison  had  been  generally  understood  before  1816.  While 
Bopp  cannot  be  credited  with  the  discovery  that  Sanskrit  is  re- 
lated to  certain  languages  of  Europe  and  Asia,  he  did  realize  that 
the  comparison  of  related  languages  could  become  the  subject 
matter  of  an  independent  science.  To  illuminate  one  language  by 
means  of  another,  to  explain  the  forms  of  one  through  the  forms 
of  the  other,  that  is  what  no  one  had  done  before  him. 

Whether  Bopp  could  have  created  his  science — so  quickly  at 
least — without  the  prior  discovery  of  Sanskrit  is  doubtful.  With 
Sanskrit  as  a  third  witness  beside  Latin  and  Greek,  Bopp  had  a 
larger  and  firmer  basis  for  his  studies.  Fortunately,  Sanskrit  was 
exceptionally  well-fitted  to  the  role  of  illuminating  the  comparison. 

For  example,  a  comparison  of  the  paradigms  of  Latin  genus 
(genus,  generis,  genere,  genera,  generum,  etc.)  and  Greek  (genos, 
gineos,  genei,  genea,  geneon,  etc.)  reveals  nothing.  But  the  picture 
changes  as  soon  as  we  add  the  corresponding  Sanskrit  series  (ganas, 
ganasas,  ganasi,  ganasu,  ^anasdm,  etc.).  A  glance  reveals  the  simi- 
larity between  the  Greek  forms  and  the  Latin  forms.  If  we  ac- 
cept tentatively  the  hypothesis  that  ^anas  represents  the  primi- 
tive state — and  this  step  facilitates  explanation — then  we  conclude 
that  s  must  have  fallen  in  Greek  forms  wherever  it  occurred  be- 
tween two  vowels.  Next  we  conclude  that  s  became  r  in  Latin  under 
the  same  conditions.  Grammatically,  then,  the  Sanskrit  paradigm 
exemplifies  the  concept  of  radical,  a  unit  (ganas)  that  is  quite 
definite  and  stable.  Latin  and  Greek  had  the  same  forms  as  San- 
skrit only  in  their  earlier  stages.  Here  Sanskrit  is  instructive  pre- 
cisely because  it  has  preserved  all  the  Indo-European  s's.  Of  course 


Sanskrit  failed  in  other  respects  to  preserve  the  features  of  the 
prototype;  for  instance,  it  had  completely  revolutionized  the  vo- 
caHc  system.  But  in  general  the  original  elements  that  Sanskrit  has 
preserved  are  remarkably  helpful  in  research — and  fate  decreed 
that  it  was  to  clarify  many  points  in  the  study  of  other  languages. 

Other  distinguished  linguists  soon  added  to  the  contribution  of 
Bopp :  Jacob  Grimm,  the  founder  of  Germanic  studies  (his  Deutsche 
Grammatik  was  published  from  1822  to  1836) ;  Pott,  whose  etymo- 
logical studies  made  a  considerable  amount  of  material  available 
to  linguists;  Kuhn,  whose  works  dealt  with  both  linguistics 
and  comparative  mythology;  the  Indie  scholars  Benfey  and 
Aufrecht,  etc. 

Finally,  among  the  last  representatives  of  the  school,  Max 
Miiller,  G.  Curtius,  and  August  Schleicher  deserve  special  atten- 
tion. In  different  ways,  all  three  did  much  to  advance  comparative 
studies.  Max  Miiller  popularized  them  in  his  brilliant  discussions 
{Lessons  in  the  Science  of  Language,  1861) ;  but  his  failing  was  a 
certain  lack  of  conscientiousness.  Curtius,  a  distinguished  philol- 
ogist known  especially  for  his  Grundziige  der  griechischen  Etymologie 
(1879),  was  one  of  the  first  to  reconcile  comparative  philology  with 
classical  philology.  The  latter  had  watched  the  progress  of  the  new 
science  suspiciously,  and  each  school  had  mistrusted  the  other. 
Schleicher  was  the  first  to  try  to  codify  the  results  of  piecemeal 
investigations.  His  Compendium  der  vergleichenden  Grammatik  der 
indogermanischen  Sprachen  (1861-62)  is  more  or  less  a  systemiza- 
tion  of  the  science  founded  by  Bopp.  His  book,  with  its  long  record 
of  service,  recalls  better  than  any  other  the  broad  outlines  of  the 
comparative  school,  which  is  the  first  chapter  in  the  history  of 
Indo-European  linguistics. 

But  the  comparative  school,  which  had  the  indisputable  merit 
of  opening  up  a  new  and  fruitful  lield,  did  aot  succeed  in  setting  up 
the  true  science  of  linguistics.*  It  failed  to  seek  out  the  natureoTiIs^ 
object  of  study.  Obviously,  without  this  elementary  step,  no 
tegience  can  develop  a  method.      

The  first  mistake  t)f  the  comparative  philologists  was  also  the 
source  of  all  their  other  mistakes.  In  their  investigations  (which  em- 
braced only  the  Indo-European  languages),  they  never  asked  them- 
selves the  meaning  of  their  comparisons  or  the  significance  of  the 


relations  that  they  discovered.  Their  method  was  exclusively  com- 
parative, not  historical.  Of  course  comparison  is  required  for  any 
historical  reconstruction,  but  by  itself  it  cannot  be  conclusive.  And 
the  conclusion  was  all  the  more  elusive  whenever  the  comparative 
philologists  looked  upon  the  development  of  two  languages  as  a 
naturahst  might  look  upon  the  growth  of  two  plants.  For  example 
Schleicher,  who  always  invites  us  to  start  from  Proto-Indo-Euro- 
pean and  thus  seems  in  a  sense  to  be  a  confirmed  historian,  has  no 
hesitancy  in  saying  that  Greek  e  and  o  are  two  grades  (Stufen)  of 
the  vocalic  system.  This  is  because  Sanskrit  has  a  system  of  vocahc 
alternations  that  suggests  the  notion  of  grades.  Schleicher  supposed 
that  each  language  has  to  pass  through  those  grades  separately  and 
in  exactly  the  same  way,  just  as  plants  of  the  same  species  pass 
through  the  same  developmental  stages  independently  of  one 
another,  and  saw  a  reinforced  grade  of  e  in  Greek  o  and  a  reinforced 
grade  of  a  in  Sanskrit  a.  The  fact  is  that  a  Proto-Indo-European 
alternation  was  reflected  differently  in  Greek  and  in  Sanskrit  with- 
out there  being  any  necessary  equivalence  between  the  gram- 
matical effects  produced  in  either  language  (see  pp.  158  ff.). 

The  exclusively  comparative  method  brought  in  a  set  of  false 
notions.  Having  no  basis  in  reality,  these  notions  simply  could  not 
reflect  the  facts  of  speech.  Language  was  considered  a  specific 
sphere,  a  fourth  natural  kingdom ;  this  led  to  methods  of  reasoning 
which  would  have  caused  astonishment  in  other  sciences.  Today 
one  cannot  read  a  dozen  lines  written  at  that  time  without  being 
struck  by  absurdities  of  reasoning  and  by  the  terminology  used 
to  justify  these  absurdities. 

But  from  the  viewpoint  of  methodology,  the  mistakes  of  the 
comparative  philologists  are  not  without  value;  the  mistakes  of  an 
infant  science  give  a  magnified  picture  of  those  made  by  anyone  in 
the  first  stages  of  scientific  research,  and  I  shall  have  occasion  to 
point  out  several  of  them  in  the  course  of  this  exposition. 

Not  until  around  1870  did  scholars  begin  to  seek  out  the  prin- 
ciples that  govern  the  life  of  languages.  Then  they  began  to  see 
that  similarities  between  languages  are  only  one  side  of  the  lin- 
guistic phenomenon,  that  comparison  is  only  a  means  or  method  of 
reconstructing  the  facts. 

Linguistics  proper,  which  puts  comparative  studies  in  their 


proper  place,  owes  its  origin  to  the  study  of  the  Romance  and 
Germanic  languages.  Romance  studies,  begun  by  Diez — his  Gram- 
matik  der  romanischen  Sprachen  dates  from  1836-38 — were  in- 
strumental in  bringing  linguistics  nearer  to  its  true  object.  For 
Romance  scholars  enjoyed  privileged  conditions  that  were  un- 
known to  Indo-European  scholars.  They  had  direct  access  to  Latin, 
the  prototype  of  the  Romance  languages,  and  an  abundance  of 
texts  allowed  them  to  trace  in  detail  the  evolution  of  the  different 
dialects;  these  two  circumstances  narrowed  the  field  of  conjecture 
and  provided  a  remarkably  solid  frame  for  all  their  research. 
Germanic  scholars  were  in  a  similar  situation.  Though  they  could 
not  study  the  prototype  directly,  numerous  texts  enabled  them  to 
trace  the  history  of  the  languages  derived  from  Proto-Germanic 
through  the  course  of  many  centuries.  The  Germanic  scholars, 
coming  to  closer  grips  with  reality  than  had  the  first  Indo-Euro- 
pean scholars,  reached  different  conclusions. 

A  first  impetus  was  given  by  the  American  scholar  Whitney,  the 
author  of  Life  and  Growth  of  Language  (1875).  Shortly  afterwards 
a  new  school  was  formed  by  the  neogrammarians  (Junggram- 
matiker),  whose  leaders  were  all  Germans:  K.  Brugmann  and  H. 
Osthoff;  the  Germanic  scholars  W.  Braune,  E.  Sievers,  H.  Paul; 
the  Slavic  scholar  Leskien,  etc.  Their  contribution  was  in  placing 
the  results  of  comparative  studies  in  their  historical  perspective- , 
and  thus  linking  the  facts  in  their  natural  order.- Thanks  to  them, 
language  is  no  longer  looked  upon  as  an  organism  that  develops 
independently  but  as  a  product  of  the  collective  mind  of  linguistic 
,  groups.yA't  the  same  time  scholars  realized  how  erroneous  and  in- 
Tsufficient  were  the  notions  of  philology  and  comparative  philology.^ 
Still,  in  spite  of  the  services  that  they  rendered,  the  neogram- 
marians did  not  illuminate  the  whole  question,  and  the  funda- 
mental problems  of  general  linguistics  still  await  solution. 

*  The  new  school,  using  a  more  reahstic  approach  than  had  its  predecessor, 
fought  the  terminology  of  the  comparative  school,  and  especially  the  illogical 
metaphors  that  it  used.  One  no  longer  dared  to  say,  "Language  does  this  or 
that,"  or  "life  of  language,"  etc.  since  language  is  not  an  entity  and  exists 
only  within  speakers.  One  must  not  go  too  far,  however,  and  a  compromise 
is  in  order.  Certain  metaphors  are  indispensable.  To  require  that  only  words 
that  correspond  to  the  facts  of  speech  be  used  is  to  pretend  that  these  facts 
no  longer  perplex  us.  This  is  by  no  means  true,  and  in  some  instances  I  shall 
not  hesitate  to  use  one  of  the  expressions  condemned  at  that  time.  [S.] 


Chapter  II 


The  subject  matter  of  linguistics  comprises  all  manifestations  of 
human  speech,  whether  that  of  savages  or  civilized  nations,  or  of 
archaic,  classical  or  decadent  periods.  In  each  period  the  linguist 
must  consider  not  only  correct  speech  and  flowery  language,  but  all 
other  forms  of  expression  as  well.  And  that  is  not  all:  since  he  is 
often  unable  to  observe  speech  directly,  he  must  consider  written 
texts,  for  only  through  them  can  he  reach  idioms  that  are  remote 
in  time  or  space. 

The  scope  of  linguistics  should  be : 

a)  to  describe  and  trace  the  history  of  all  observable  languages, 
which  amounts  to  tracing  the  history  of  families  of  languages  and 
reconstructing  as  far  as  possible  the  mother  language  of  each 

6)  to  determine  the  forces  that  are  permanently  and  universally 
at  work  in  all  languages,  and  to  deduce  the  general  laws  to  which 
all  specific  historical  phenomena  can  be  reduced;  arid 

c)   to  delimit  and  define  itself. 

Linguistics  is  very  closely  related  to  other  sciences  that  some- 
times borrow  from  its  data,  sometimes  supply  it  with  data.  The 
lines  of  demarcation  do  not  always  show  up  clearly.  For  instance, 
linguistics  must  be  carefully  distinguished  from  ethnography  and 
prehistory,  where  language  is  used  merely  to  document.  It  must 
also  be  set  apart  from  anthropology,  which  studies  man  solely  from 
the  viewpoint  of  his  species,  for  language  is  a  social  fact.  But  must 
linguistics  then  be  combined  with  sociology?  What  are  the  relation- 
ships between  linguistics  and  social  psychology?  Everything  in 
language  is  basically  psychological,  including  its  material  and 
mechanical  manifestations,  such  as  sound  changes;  and  since  lin- 
guistics provides  social  psychology  with  such  valuable  data,  is  it 


not  part  and  parcel  of  this  discipline?  Here  I  shall  raise  many  sim- 
ilar questions ;  later  I  shall  treat  them  at  greater  length. 

The  ties  between  linguistics  and  the  physiology  of  sounds  are 
less  difficult  to  untangle.  The  relation  is  unilateral  in  the  sense  that 
the  study  of  languages  exacts  clarifications  from  the  science  of  the 
physiology  of  sounds  but  furnishes  none  in  return.  In  any  event, 
the  two  disciplines  cannot  be  confused.  The  thing  that  constitutes 
language  is,  as  I  shall  show  later,  unrelated  to  the  phonic  character 
of  the  linguistic  sign. 

As  for  philology,  we  have  already  drawn  the  line:  it  is  distinct 
from  linguistics  despite  points  of  contact  between  the  two  sciences 
and  mutual  services  that  they  render. 

Finally,  of  what  use  is  linguistics?  Very  few  people  have  clear 
ideas  on  this  point,  and  this  is  not  the  place  to  specify  them.  But  it 
is  evident,  for  instance,  that  linguistic  questions  interest  all  who 
work  with  texts — historians,  philologists,  etc.  Still  more  obvious  is 
the  importance  of  linguistics  to  general  culture:  in  the  lives  of 
individuals  and  societies,  speech  is  more  important  than  anything 
else.  That  linguistics  should  continue  to  be  the  prerogative  of  a  few 
specialists  would  be  unthinkable — everyone  is  concerned  with  it  in 
one  way  or  another.  But — and  this  is  a  paradoxical  consequence  of 
the  interest  that  is  fixed  on  hnguistics — there  is  no  other  field  in 
which  so  many  absurd  notions,  prejudices,  mirages,  and  fictions 
have  sprung  up.  From  the  psychological  viewpoint  these  errors 
are  of  interest,  but  the  task  of  the  linguist  is,  above  all  else,  to 
condemn  them  and  to  dispel  them  as  best  he  can. 

Chapter  III 


1.  Definition  of  Language 

What  is  both  the  integral  and  concrete  object  of  linguistics?  The 
question  is  especially  difficult;  later  we  shall  see  why;  here  I  wish 
merely  to  point  up  the  difficulty. 


Other  sciences  work  with  objects  that  are  given  in  advance  and 
that  can  then  be  considered  from  different  viewpoints;  but  not 
Hnguistics.  Someone  pronounces  the  French  word  nu  'bare':  a 
superficial  observer  would  be  tempted  to  call  the  word  a  concrete 
linguistic  object;  but  a  more  careful  examination  would  reveal 
successively  three  or  four  quite  different  things,  depending  on 
whether  the  word  is  considered  as  a  sound,  as  the  expression  of  an 
idea,  as  the  equivalent  of  Latin  nudum,  etc.  Far  from  it  being  the 
object  that  antedates  the  viewpoint,  it  would  seem  that  it  is  the 
viewpoint  that  creates  the  object;  besides,  nothing  tells  us  in 
advance  that  one  way  of  considering  the  fact  in  question  takes 
precedence  over  the  others  or  is  in  any  way  superior  to  them. 

Moreover,  regardless  of  the  viewpoint  that  we  adopt,  the  lin- 
guistic phenomenon  always  has  two  related  sides,  each  deriving  its 
values  from  the  other.  For  example : 

1)  Articulated  syllables  are  acoustical  impressions  perceived  by 
the  ear,  but  the  sounds  would  not  exist  without  the  vocal  organs ; 
an  n,  for  example,  exists  only  by  virtue  of  the  relation  between  the 
two  sides.  We  simply  cannot  reduce  language  to  sound  or  detach 
sound  from  oral  articulation;  reciprocally,  we  cannot  define  the 
movements  of  the  vocal  organs  without  taking  into  account  the 
acoustical  impression  (see  pp.  38  ff.). 

2)  But  suppose  that  sound  were  a  simple  thing:  would  it  consti- 
tute speech?  No,  it  is  only  the  instrument  of  thought;  by  itself,  it 
has  no  existence.  At  this  point  a  new  and  redoubtable  relationship 
arises:  a  sound,  a  complex  acoustical-vocal  unit,  combines  in  turn 
with  an  idea  to  form  a  complex  physiological-psychological  unit. 
But  that  is  still  not  the  complete  picture. 

3)  Speech  has  both  an  individual  and  a  social  side,  and  we  can- 
not conceive  of  one  without  the  other.  Besides : 

4)  Speech  always  implies  both  an  established  system  and  an 
evolution;  at  every  moment  it  is  an  existing  institution  and  a 
product  of  the  past.  To  distinguish  between  the  system  and  its 
history,  between  what  it  is  and  what  it  was,  seems  very  simple  at 
first  glance ;  actually  the  two  things  are  so  closely  related  that  we 
can  scarcely  keep  them  apart.  Would  we  simplify  the  question  by 
studying  the  linguistic  phenomenon  in  its  earliest  stages — if  we 


began,  for  example,  by  studying  the  speech  of  children?  No,  for  in 
dealing  with  speech,  it  is  completely  misleading  to  assume  that  the 
problem  of  early  characteristics  differs  from  the  problem  of  per- 
manent characteristics.  We  are  left  inside  the  vicious  circle. 

From  whatever  direction  we  approach  the  question,  nowhere  do 
we  find  the  integral  object  of  linguistics.  Everywhere  we  are  con- 
fronted with  a  dilemma :  if  we  fix  our  attention  on  only  one  side  of 
each  problem,  we  run  the  risk  of  failing  to  perceive  the  dualities 
pointed  out  above;  on  the  other  hand,  if  we  study  speech  from 
several  viewpoints  simultaneously,  the  object  of  linguistics  appears 
to  us  as  a  confused  mass  of  heterogeneous  and  unrelated  things. 
Either  procedure  opens  the  door  to  several  sciences — psychology, 
anthropology,  normative  grammar,  philology,  etc. — which  are 
distinct  from  linguistics,  but  which  might  claim  speech,  in  view  of 
the  faulty  method  of  linguistics,  as  one  of  their  objects. 

As  I  see  it  there  is  only  one  solution  to  all  the  foregoing  difl5- 
culties :  from  the  very  outset  we  must  put  both  feet  on  the  ground  of 
language  and  use  language  as  the  norm  of  all  other  manifestations  of 
speech.  Actually,  among  so  many  dualities,  language  alone  seems 
to  lend  itself  to  independent  definition  and  provide  a  fulcrum  that 
satisfies  the  mind. 

But  what  is  language  [Zangwe]?  It  is  not  to  be  confused  with 
human  speech  [langage],  of  which  it  is  only  a  definite  part,  though 
certainly  an  essential  one.  It  is  both  a  social  product  of  the  faculty 
of  speech  and  a  collection  of  necessary  conventions  that  have  been 
adopted  by  a  social  body  to  permit  individuals  to  exercise  that 
faculty.  Taken  as  a  whole,  speech  is  many-sided  and  heterogene- 
ous; straddling  several  areas  simultaneously — physical,  physio- 
logical, and  psychological — it  belongs  both  to  the  individual  and 
to  society ;  we  cannot  put  it  into  any  category  of  human  facts,  for 
we  cannot  discover  its  unity. 

Language,  on  the  contrary,  is  a  seK-contained  whole  and  a  prin- 
ciple of  classification.  As  soon  as  we  give  language  first  place  among 
the  facts  of  speech,  we  introduce  a  natural  order  into  a  mass  that 
lends  itself  to  no  other  classification. 

One  might  object  to  that  principle  of  classification  on  the  ground 
that  since  the  use  of  speech  is  based  on  a  natural  faculty  whereas 


language  is  something  acquired  and  conventional,  language  should 
not  take  first  place  but  should  be  subordinated  to  the  natural 

That  objection  is  easily  refuted. 

First,  no  one  has  proved  that  speech,  as  it  manifests  itself  when 
we  speak,  is  entirely  natural,  i.e.  that  our  vocal  apparatus  was 
designed  for  speaking  just  as  our  legs  were  designed  for  walking. 
Linguists  are  far  from  agreement  on  this  point.  For  instance  Whit- 
ney, to  whom  language  is  one  of  several  social  institutions,  thinks 
that  we  use  the  vocal  apparatus  as  the  instrument  of  language 
purely  through  luck,  for  the  sake  of  convenience:  men  might  just 
as  well  have  chosen  gestures  and  used  visual  symbols  instead  of 
acoustical  symbols.  Doubtless  his  thesis  is  too  dogmatic ;  language 
is  not  similar  in  all  respects  to  other  social  institutions  (see  p.  73  f . 
and  p.  75  f.);  moreover,  Whitney  goes  too  far  in  saying  that  our 
choice  happened  to  fall  on  the  vocal  organs;  the  choice  was  more 
or  less  imposed  by  nature.  But  on  the  essential  point  the  American 
linguist  is  right:  language  is  a  convention,  and  the  nature  of  the 
sign  that  is  agreed  upon  does  not  matter.  The  question  of  the  vocal 
apparatus  obviously  takes  a  secondary  place  in  the  problem  of 

One  definition  of  articulated  speech  might  confirm  that  conclusion. 
In  Latin,  articulus  means  a  member,  part,  or  subdivision  of  a 
sequence ;  applied  to  speech,  articulation  designates  either  the  sub- 
division of  a  spoken  chain  into  syllables  or  the  subdivision  of  the 
chain  of  meanings  into  significant  units ;  gegliederte  Sprache  is  used 
in  the  second  sense  in  German.  Using  the  second  definition,  we  can 
say  that  what  is  natural  to  mankind  is  not  oral  speech  but  the 
faculty  of  constructing  a  language,  i.e.  a  system  of  distinct  signs 
corresponding  to  distinct  ideas. 

Broca  discovered  that  the  faculty  of  speech  is  localized  in  the 
third  left  frontal  convolution ;  his  discovery  has  been  used  to  sub- 
stantiate the  attribution  of  a  natural  quality  to  speech.  But  we 
know  that  the  same  part  of  the  brain  is  the  center  of  everything  that 
has  to  do  with  speech,  including  writing.  The  preceding  statements, 
together  with  observations  that  have  been  made  in  different  cases 
of  aphasia  resulting  from  lesion  of  the  centers  of  localization,  seem 
to  indicate:  (1)  that  the  various  disorders  of  oral  speech  are  bound 


up  in  a  hundred  ways  with  those  of  written  speech;  and  (2)  that 
what  is  lost  in  all  cases  of  aphasia  or  agraphia  is  less  the  faculty  of 
producing  a  given  sound  or  writing  a  given  sign  than  the  ability  to 
evoke  by  means  of  an  instrument,  regardless  of  what  it  is,  the  signs 
of  a  regular  system  of  speech.  The  obvious  implication  is  that 
beyond  the  functioning  of  the  various  organs  there  exists  a  more 
general  faculty  which  governs  signs  and  which  would  be  the 
linguistic  faculty  proper.  And  this  brings  us  to  the  same  conclusion 
as  above. 

To  give  language  first  place  in  the  study  of  speech,  we  can  ad- 
vance a  final  argument :  the  faculty  of  articulating  words — ^whether 
it  is  natural  or  not — is  exercised  only  with  the  help  of  the  instru- 
ment created  by  a  collectivity  and  provided  for  its  use;  therefore, 
to  say  that  language  gives  unity  to  speech  is  not  fanciful. 

2.  Place  of  Language  in  the  Facts  of  Speech 

In  order  to  separate  from  the  whole  of  speech  the  part  that  be- 
longs to  language,  we  must  examine  the  individual  act  from  which 
the  speaking-circuit  can  be  reconstructed.  The  act  requires  the 
presence  of  at  least  two  persons;  that  is  the  minimum  number 
necessary  to  complete  the  circuit.  Suppose  that  two  people,  A  and 
B,  are  conversing  with  each  other : 

»•- n^,    j^jjir-'* 

Suppose  that  the  opening  of  the  circuit  is  in  A's  brain,  where 
mental  facts  (concepts)  are  associated  with  representations  of  the 
linguistic  sounds  (sound-images)  that  are  used  for  their  expression. 
A  given  concept  unlocks  a  corresponding  sound-image  in  the  brain ; 
this  purely  psychological  phenomenon  is  followed  in  turn  by  a 
physiological  process :  the  brain  transmits  an  impulse  corresponding 


to  the  image  to  the  organs  used  in  producing  sounds.  Then  the 
sound  waves  travel  from  the  mouth  of  A  to  the  ear  of  B :  a  purely 
physical  process.  Next,  the  circuit  continues  in  B,  but  the  order  is 
reversed:  from  the  ear  to  the  brain,  the  physiological  transmission 
of  the  sound-image;  in  the  brain,  the  psychological  association  of 
the  image  with  the  corresponding  concept.  If  B  then  speaks,  the 
new  act  will  follow — from  his  brain  to  A's — exactly  the  same  course 
as  the  first  act  and  pass  through  the  same  successive  phases,  which 
I  shall  diagram  as  follows : 

Audition  Phonatlon 

«C    < 

c  =  concept 

s  =z  sound-Image 


Phonation  Audition 

The  preceding  analysis  does  not  purport  to  be  complete.  We 
might  also  single  out  the  pure  acoustical  sensation,  the  identifi- 
cation of  that  sensation  with  the  latent  sound-image,  the  muscular 
image  of  phonation,  etc.  I  have  included  only  the  elements  thought 
to  be  essential,  but  the  drawing  brings  out  at  a  glance  the  distinc- 
tion between  the  physical  (sound  waves),  physiological  (phonation 
and  audition),  and  psychological  parts  (word-images  and  con- 
cepts). Indeed,  we  should  not  fail  to  note  that  the  word-image 
stands  apart  from  the  sound  itself  and  that  it  is  just  as  psycho- 
logical as  the  concept  which  is  associated  with  it. 

The  circuit  that  I  have  outlined  can  be  further  divided  into: 

a)  an  outer  part  that  includes  the  vibrations  of  the  sounds  which 
travel  from  the  mouth  to  the  ear,  and  an  inner  part  that  includes 
everything  else ; 

h)  a  psychological  and  a  nonpsychological  part,  the  second  in- 
cluding the  physiological  productions  of  the  vocal  organs  as  well 
as  the  physical  facts  that  are  outside  the  individual ; 


c)  an  active  and  a  passive  part:  everything  that  goes  from  the 
associative  center  of  the  speaker  to  the  ear  of  the  hstener  is  active, 
and  everything  that  goes  from  the  ear  of  the  hstener  to  his  associ- 
ative center  is  passive; 

d)  finally,  everything  that  is  active  in  the  psychological  part  of 
the  circuit  is  executive  {c  -^  s),  and  everything  that  is  passive  is 
receptive  (s  — >  c). 

We  should  also  add  the  associative  and  co-ordinating  faculty 
that  we  find  as  soon  as  we  leave  isolated  signs;  this  faculty  plays 
the  dominant  role  in  the  organization  of  language  as  a  system  (see 
pp.  122  ff.)- 

But  to  understand  clearly  the  role  of  the  associative  and  co- 
ordinating faculty,  we  must  leave  the  individual  act,  which  is  only 
the  embryo  of  speech,  and  approach  the  social  fact. 

Among  all  the  individuals  that  are  linked  together  by  speech, 
some  sort  of  average  will  be  set  up :  all  will  reproduce — not  exactly 
of  course,  but  approximately — the  same  signs  united  with  the 
same  concepts. 

How  does  the  social  crystallization  of  language  come  about? 
Which  parts  of  the  circuit  are  involved?  For  all  parts  probably  do 
not  participate  equally  in  it. 

The  nonpsychological  part  can  be  rejected  from  the  outset. 
When  we  hear  people  speaking  a  language  that  we  do  not  know, 
we  perceive  the  sounds  but  remain  outside  the  social  fact  because 
we  do  not  understand  them. 

Neither  is  the  psychological  part  of  the  circuit  wholly  respon- 
sible: the  executive  side  is  missing,  for  execution  is  never  carried 
out  by  the  collectivity.  Execution  is  always  individual,  and  the 
individual  is  always  its  master:  I  shall  call  the  executive  side 
speaking  [parole]. 

Through  the  functioning  of  the  receptive  and  co-ordinating 
faculties,  impressions  that  are  perceptibly  the  same  for  all  are  made 
on  the  minds  of  speakers.  How  can  that  social  product  be  pictured 
in  such  a  way  that  language  will  stand  apart  from  everything  else? 
If  we  could  embrace  the  sum  of  word-images  stored  in  the  minds 
of  all  individuals,  we  could  identify  the  social  bond  that  consti- 
tutes language.  It  is  a  storehouse  filled  by  the  members  of  a  given 
community  through  their  active  use  of  speaking,  a  grammatical 


system  that  has  a  potential  existence  in  each  brain,  or,  more 
specifically,  in  the  brains  of  a  group  of  individuals.  For  language 
is  not  complete  in  any  speaker;  it  exists  perfectly  only  within  a 

In  separating  language  from  speaking  we  are  at  the  same  time 
separating:  (1)  what  is  social  from  what  is  individual;  and  (2)  what 
is  essential  from  what  is  accessory  and  more  or  less  accidental. 

Language  is  not  a  function  of  the  speaker ;  it  is  a  product  that  is 
passively  assimilated  by  the  individual.  It  never  requires  premedi- 
tation, and  reflection  enters  in  only  for  the  purpose  of  classification, 
which  we  shall  take  up  later  (pp.  122  ff.). 

— Speaking,  on  the  contrary,  is  an  individual  act.  It  is  wilful  and 
intellectual.  Within  the  act,  we  should  distinguish  between:  (1)  the 
combinations  by  which  the  speaker  uses  the  language  code  for 
expressing  his  own  thought;  and  (2)  the  psychophysical  mecha- 
nism that  allows  him  to  exteriorize  those  combinations. 

Note  that  I  have  defined  things  rather  than  words ;  these  defini- 
tions are  not  endangered  by  certain  ambiguous  words  that  do  not 
have  identical  meanings  in  different  languages.  For  instance, 
German  Sprache  means  both  "language"  and  "speech";  Rede 
almost  corresponds  to  "speaking"  but  adds  the  special  connotation 
of  "discourse."  Latin  sermo  designates  both  "speech"  and  "speak- 
ing," while  lingua  means  "language,"  etc.  No  word  corresponds 
exactly  to  any  of  the  notions  specified  above ;  that  is  why  all  defini- 
tions of  words  are  made  in  vain;  starting  from  words  in  defining 
things  is  a  bad  procedure. 

To  summarize,  these  are  the  characteristics  of  language : 

1)  Language  is  a  well-defined  object  in  the  heterogeneous  mass 
of  speech  facts.  It  can  be  localized  in  the  limited  segment  of  the 
speaking-circuit  where  an  auditory  image  becomes  associated  with 
a  concept.  It  is  the  social  side  of  speech,  outside  the  individual  who 
can  never  create  nor  modify  it  by  himself;  it  exists  only  by  virtue 
of  a  sort  of  contract  signed  by  the  members  of  a  community.  More- 
over, the  individual  must  always  serve  an  apprenticeship  in  order 
to  learn  the  functioning  of  language;  a  child  assimilates  it  only 
gradually.  It  is  such  a  distinct  thing  that  a  man  deprived  of  the 
use  of  speaking  retains  it  provided  that  he  understands  the  vocal 
signs  that  he  hears. 


2)  Language,  unlike  speaking,  is  something  that  we  can  study- 
separately.  Although  dead  languages  are  no  longer  spoken,  we  can 
easily  assimilate  their  linguistic  organisms.  We  can  dispense  with 
the  other  elements  of  speech;  indeed,  the  science  of  language  is 
possible  only  if  the  other  elements  are  excluded. 

3)  Whereas  speech  is  heterogeneous,  language,  as  defined,  is 
homogeneous.  It  is  a  system  of  signs  in  which  the  only  essential 
thing  is  the  union  of  meanings  and  sound-images,  and  in  which 
both  parts  of  the  sign  are  psychological. 

4)  Language  is  concrete,  no  less  so  than  speaking;  and  this  is  a 
help  in  our  study  of  it.  Linguistic  signs,  though  basically  psycho- 
logical, are  not  abstractions;  associations  which  bear  the  stamp  of 
collective  approval — and  which  added  together  constitute  language 
— are  realities  that  have  their  seat  in  the  brain.  Besides,  linguistic 
signs  are  tangible;  it  is  possible  to  reduce  them  to  conventional 
written  sjonbols,  whereas  it  would  be  impossible  to  provide  de- 
tailed photographs  of  acts  of  speaking  [actes  de  parole] ;  the  pro- 
nunciation of  even  the  smallest  word  represents  an  infinite  number 
of  muscular  movements  that  could  be  identified  and  put  into 
graphic  form  only  with  great  difficulty.  In  language,  on  the  con- 
trary, there  is  only  the  sound-image,  and  the  latter  can  be  trans- 
lated into  a  fixed  visual  image.  For  if  we  disregard  the  vast  number 
of  movements  necessary  for  the  realization  of  sound-images  in 
speaking,  we  see  that  each  sound-image  is  nothing  more  than  the 
sum  of  a  limited  number  of  elements  or  phonemes  that  can  in  turn 
be  called  up  by  a  corresponding  number  of  written  symbols  (see 
pp.  61  ff.).  The  very  possibihty  of  putting  the  things  that  relate 
to  language  into  graphic  form  allows  dictionaries  and  grammars  to 
represent  it  accurately,  for  language  is  a  storehouse  of  sound- 
images,  and  writing  is  the  tangible  form  of  those  images. 

3.  Place  of  Language  in  Human  Facts:  Semiology 

The  foregoing  characteristics  of  language  reveal  an  even  more 
important  characteristic.  Language,  once  its  boundaries  have  been 
marked  off  within  the  speech  data,  can  be  classified  among  human 
phenomena,  whereas  speech  cannot. 

We  have  just  seen  that  language  is  a  social  institution;  but  sev- 
eral features  set  it  apart  from  other  political,  legal,  etc.  institutions. 


We  must  call  in  a  new  type  of  facts  in  order  to  illuminate  the 
special  nature  of  language. 

Language  is  a  system  of  signs  that  express  ideas,  and  is  therefore 
comparable  to  a  system  of  writing,  the  alphabet  of  deaf-mutes, 
S5Tnbolic  rites,  polite  formulas,  military  signals,  etc.  But  it  is  the 
most  important  of  all  these  systems. 

A  science  that  studies  the  life  of  signs  within  society  is  conceivable ; 
it  would  be  a  part  of  social  psychology  and  consequently  of  general 
psychology;  I  shall  call  it  semiology^  (from  Greek  semeion  'sign'). 
Semiology  would  show  what  constitutes  signs,  what  laws  govern 
them.  Since  the  science  does  not  yet  exist,  no  one  can  say  what  it 
would  be ;  but  it  has  a  right  to  existence,  a  place  staked  out  in  ad- 
vance. Linguistics  is  only  a  part  of  the  general  science  of  semiology; 
the  laws  discovered  by  semiology  will  be  applicable  to  linguistics, 
and  the  latter  will  circumscribe  a  well-defined  area  within  the  mass 
of  anthropological  facts. 

To  determine  the  exact  place  of  semiology  is  the  task  of  the 
psychologist.'*  The  task  of  the  linguist  is  to  find  out  what  makes 
language  a  special  system  within  the  mass  of  semiological  data. 
This  issue  will  be  taken  up  again  later;  here  I  wish  merely  to  call 
attention  to  one  thing :  if  I  have  succeeded  in  assigning  linguistics  a 
place  among  the  sciences,  it  is  because  I  have  related  it  to  semi- 

Why  has  semiology  not  yet  been  recognized  as  an  independent 
science  with  its  own  object  like  all  the  other  sciences?  Linguists 
have  been  going  around  in  circles :  language,  better  than  anything 
else,  offers  a  basis  for  understanding  the  semiological  problem ;  but 
language  must,  to  put  it  correctly,  be  studied  in  itself;  heretofore 
language  has  almost  always  been  studied  in  connection  with  some- 
thing else,  from  other  viewpoints. 

There  is  first  of  all  the  superficial  notion  of  the  general  public : 
people  see  nothing  more  than  a  name-giving  system  in  language 
(see  p.  65),  thereby  prohibiting  any  research  into  its  true  nature. 

'  Semiology  should  not  be  confused  with  semantics,  which  studies  changes  in 
meaning,  and  which  Saussure  did  not  treat  methodically;  the  fundamental 
principle  of  semantics  is  formulated  on  page  75.  [Ed.] 

*  Cf.  A.  NaviUe,  Classification  des  Sciences,  (2nd.  ed.),  p.  104.  [Ed.]  The 
scope  of  semiology  (or  semiotics)  is  treated  at  length  in  Charles  Morris' 
Signs,  Language  and  Behavior  (New  York:  Prentice-Hall,  1946).  [Tr.] 


Then  there  is  the  viewpoint  of  the  psychologist,  who  studies  the 
sign-mechanism  in  the  individual;  this  is  the  easiest  method,  but 
it  does  not  lead  beyond  individual  execution  and  does  not  reach 
the  sign,  which  is  social. 

Or  even  when  signs  are  studied  from  a  social  viewpoint,  only  the 
traits  that  attach  language  to  the  other  social  institutions — those 
that  are  more  or  less  voluntary — are  emphasized;  as  a  result,  the 
goal  is  by-passed  and  the  specific  characteristics  of  semiological 
systems  in  general  and  of  language  in  partichlar-ace. completely 
ignored,/ I^orTFe  distinguishing  characteristic  of  the  sign— but  the"^) 
one  that  is  least  apparent  at  first  sight — is  that  in  some  way  it  '' 
arv^a.yb'  eludBSHfeheJndividual  or  .social  will. 

In  short,  the  characteristic  that  distinguishes  semiological  sys- 
tems from  all  other  institutions  shows  up  clearly  only  in  language 
where  it  manifests  itself  in  the  things  which  are  studied  least,  and 
the  necessity  or  specific  value  of  a  semiological  science  is  therefore 
not  clearly  recognized.  But  to  me  the  language  problem  is  mainly 
semiological,  and  all  developments  derive  their  significance  from 
that  important  fact.  If  we  are  to  discover  the  true  nature  of  lan- 
guage we  must  learn  what  it  has  in  common  with  all  other  semi- 
ological systems;  linguistic  forces  that  seem  very  important  at 
first  glance  (e.g.,  the  role  of  the  vocal  apparatus)  will  receive  only 
secondary  consideration  if  they  serve  only  to  set  language  apart 
from  the  other  systems.  This  procedure  will  do  more  than  to 
clarify  the  linguistic  problem.  By  studying  rites,  customs,  etc.  as 
signs,  I  believe  that  we  shall  throw  new  light  on  the  facts  and  point  / 
up  the  need  for  including  them  in  a  science  of  semiology  and 
explaining  them  by  its  laws. 

Chapter  IV 


In  setting  up  the  science  of  language  within  the  overall  study  of 
speech,  I  have  also  outlined  the  whole  of  linguistics.  All  other  ele- 


merits  of  speech — those  that  constitute  speaking — freely  subordi- 
nate themselves  to  the  first  science,  and  it  is  by  virtue  of  this  sub- 
ordination that  the  parts  of  linguistics  find  their  natural  place. 

Consider,  for  example,  the  production  of  sounds  necessary  for 
speaking.  The  vocal  organs  are  as  external  to  language  as  are  the 
electrical  devices  used  in  transmitting  the  Morse  code  to  the  code 
itself;  and  phonation,  i.e.,  the  execution  of  sound-images,  in  no  way 
affects  the  system  itself.  Language  is  comparable  to  a  symphony 
in  that  what  the  sjnnphony  actually  is  stands  completely  apart 
from  how  it  is  performed;  the  mistakes  that  musicians  make  in 
playing  the  symphony  do  not  compromise  this  fact. 

An  argument  against  separating  phonation  from  language  might 
be  phonetic  changes,  the  alterations  of  the  sounds  which  occur  in 
speaking  and  which  exert  such  a  profound  influence  on  the  future 
of  language  itself.  Do  we  really  have  the  right  to  pretend  that  lan- 
guage exists  independently  of  phonetic  changes?  Yes,  for  they 
affect  only  the  material  substance  of  words.  If  they  attack  language 
as  a  system  of  signs,  it  is  only  indirectly,  through  subsequent 
changes  of  interpretation;  there  is  nothing  phonetic  in  the  phe- 
nomenon (see  p.  84).  Determining  the  causes  of  phonetic  changes 
may  be  of  interest,  and  the  study  of  sounds  will  be  helpful  on  this 
point ;  but  none  of  this  is  essential :  in  the  science  of  language,  all 
we  need  do  is  to  observe  the  transformations  of  sounds  and  to 
calculate  their  effects. 

What  I  have  said  about  phonation  applies  to  all  other  parts  of 
speaking.  The  activity  of  the  speaker  should  be  studied  in  a  num- 
ber of  disciplines  which  have  no  place  in  linguistics  except  through 
their  relation  to  language. 

The  study  of  speech  is  then  twofold :  its  basic  part — ^having  as  its 
object  language,  which  is  purely  social  and  independent  of  the 
individual — is  exclusively  psychological ;  its  secondary  part — which 
has  as  its  object  the  individual  side  of  speech,  i.e.  speaking,  includ- 
ing phonation — is  psychophysical. 

Doubtless  the  two  objects  are  closely  connected,  each  depending 
on  the  other :  language  is  necessary  if  speaking  is  to  be  intelligible 
and  produce  all  its  effects;  but  speaking  is  necessary  for  the  estab- 
lishment of  language,  and  historically  its  actuality  always  comes 
first.  How  would  a  speaker  take  it  upon  himself  to  associate  an  idea 


with  a  word-image  if  he  had  not  first  come  across  the  association  in 
an  act  of  speaking?  Moreover,  we  learn  our  mother  language  by 
listening  to  others;  only  after  countless  experiences  is  it  deposited 
in  our  brain.  Finally,  speaking  is  what  causes  language  to  evolve: 
impressions  gathered  from  listening  to  others  modify  our  linguistic 
habits.  Language  and  speaking  are  then  interdependent ;  the  former 
is  both  the  instrument  and  the  product  of  the  latter.  But  their 
interdependence  does  not  prevent  their  being  two  absolutely 
distinct  things. 

Language  exists  in  the  form  of  a  sum  of  impressions  deposited  in 
the  brain  of  each  member  of  a  community,  almost  like  a  dictionary 
of  which  identical  copies  have  been  distributed  to  each  individual 
(see  p.  13).  Language  exists  in  each  individual,  yet  is  common  to 
all.  Nor  is  it  affected  by  the  will  of  the  depositaries.  Its  mode  of 
existence  is  expressed  by  the  formula: 

1  +  1  +  1  +  1...  =  1  (collective  pattern) 

What  part  does  speaking  play  in  the  same  community?  It  is  the 
sum  of  what  people  say  and  includes :  (a)  individual  combinations 
that  depend  on  the  will  of  speakers,  and  (b)  equally  wilful  pho- 
national  acts  that  are  necessary  for  the  execution  of  these  com- 

Speaking  is  thus  not  a  collective  instrument;  its  manifestations 
are  individual  and  momentary.  In  speaking  there  is  only  the  sum  of 
particular  acts,  as  in  the  formula : 

(1  +  r  +  1"  +  1'".  .  .) 

For  all  the  foregoing  reasons,  to  consider  language  and  speaking 
from  the  same  viewpoint  would  be  fanciful.  Taken  as  a  whole, 
speech  cannot  be  studied,  for  it  is  not  homogeneous;  but  the  dis- 
tinction and  subordination  proposed  here  clarify  the  whole  issue. 

Such  is  the  first  bifurcation  that  we  find  in  trying  to  formulate 
the  theory  of  speech.  We  must  choose  between  two  routes  that 
cannot  be  followed  simultaneously;  they  must  be  followed 

One  might  if  really  necessary  apply  the  term  linguistics  to  each 
of  the  two  disciplines  and  speak  of  a  linguistics  of  speaking.  But 


that  science  must  not  be  confused  with  Hnguistics  proper,  whose 
sole  object  is  language. 

I  shall  deal  only  with  linguistics  of  language,  and  if  I  sub- 
sequently use  material  belonging  to  speaking  to  illustrate  a  point, 
I  shall  try  never  to  erase  the  boundaries  that  separate  the  two 

Chapter  V 


My  definition  of  language  presupposes  the  exclusion  of  everything 
that  is  outside  its  organism  or  system — in  a  word,  of  everything 
known  as  "external  linguistics."  But  external  linguistics  deals  with 
many  important  things — the  very  ones  that  we  think  of  when  we 
begin  the  study  of  speech. 

First  and  foremost  come  all  the  points  where  linguistics  borders 
on  ethnology,  all  the  relations  that  link  the  history  of  a  language 
and  the  history  of  a  race  or  civilization.  The  close  interaction  of 
language  and  ethnography  brings  to  mind  the  bonds  that  join  lin- 
guistic phenomena  proper  (see  pp.  7  f.).  The  culture  of  a  nation 
exerts  an  influence  on  its  language,  and  the  language,  on  the  other 
hand,  is  largely  responsible  for  the  nation. 

Second  come  the  relations  between  language  and  political  his- 
tory. Great  historical  events  like  the  Roman  conquest  have  an 
incalculable  influence  on  a  host  of  hnguistic  facts.  Colonization, 
which  is  only  one  form  that  conquest  may  take,  brings  about 
changes  in  an  idiom  by  transporting  it  into  different  surroundings. 
All  kinds  of  facts  could  be  cited  as  substantiating  evidence.  For 
instance,  Norway  adopted  Danish  when  she  united  politically  with 
Denmark;  the  Norwegians  are  trying  today  to  throw  off  that 
linguistic  influence.  The  internal  politics  of  states  is  no  less  im- 
portant to  the  life  of  languages;  certain  governments  (Uke  the 
Swiss)  allow  the  coexistence  of  several  idioms;  others  (like  the 
French)  strive  for  linguistic  unity.  An  advanced  state  of  civihzation 


favors  the  development  of  special  languages  (juridical  language, 
scientific  terminology,  etc.). 

Here  we  come  to  a  third  point:  the  relations  between  language 
and  all  sorts  of  institutions  (the  Church,  the  school,  etc.).  All  these 
institutions  in  turn  are  closely  tied  to  the  literary  development  of 
a  language,  a  general  phenomenon  that  is  all  the  more  inseparable 
from  political  history.  At  every  point  the  literary  language  over- 
steps the  boundaries  that  literature  apparently  marks  off;  we  need 
only  consider  the  influence  of  salons,  the  court,  and  national 
academies.  Moreover,  the  literary  language  raises  the  important 
question  of  conflicts  between  it  and  local  dialects  (see  pp.  195  ff.); 
the  linguist  must  also  examine  the  reciprocal  relations  of  book 
language  and  the  vernacular;  for  every  literary  language,  being  the 
product  of  the  culture,  finally  breaks  away  from  its  natural  sphere, 
the  spoken  language. 

Finally,  everything  that  relates  to  the  geographical  spreading  of 
languages  and  dialectal  splitting  belongs  to  external  linguistics. 
Doubtless  the  distinction  between  internal  and  external  linguistics 
seems  most  paradoxical  here,  since  the  geographical  phenomenon 
is  so  closely  linked  to  the  existence  of  any  language ;  but  geographi- 
cal spreading  and  dialectal  splitting  do  not  actually  affect  the  inner 
organism  of  an  idiom. 

Some  have  maintained  that  the  foregoing  issues  simply  cannot 
be  separated  from  the  study  of  language  proper.  The  viewpoint 
has  been  prevalent  especially  since  the  placing  of  so  much  emphasis 
on  "Realia."^  Just  as  the  inner  organism  of  a  plant  is  modified  by 
alien  forces  (terrain,  cUmate,  etc.)  does  not  the  grammatical 
organism  depend  constantly  on  the  external  forces  of  linguistic 
change?  It  seems  that  we  can  scarcely  give  a  satisfactory  expla- 
nation of  the  technical  terms  and  loan-words  that  abound  in  lan- 
guage without  considering  their  development.  Is  it  possible  to 
distinguish  the  natural,  organic  growth  of  an  idiom  from  its  arti- 
ficial forms,  such  as  the  literary  language,  which  are  due  to  ex- 
ternal, and  therefore  inorganic  forces?  Common  languages  are 
always  developing  alongside  local  dialects, 

^  Realien  is  used  in  German  to  refer  to  all  material  facts  of  life,  the  shape, 
dimensions,  and  the  like  of  objects,  things,  etc.  Cf.  the  numerous  works  in 
German  entitled  Reallexicon.  [Tr.] 


I  believe  that  the  study  of  external  linguistic  phenomena  is  most 
fruitful;  but  to  say  that  we  cannot  understand  the  internal  lin- 
guistic organism  without  studying  external  phenomena  is  wrong. 
Take  as  an  example  the  borrowing  of  foreign  words.  We  observe 
from  the  outset  that  borrowing  is  not  a  constant  force  in  the  life  of 
a  language.  In  certain  isolated  valleys  there  are  dialects  that  have 
never  taken  a  single  artificial  term  from  the  outside.  Should  we  say 
that  such  idioms  are  outside  the  conditions  of  normal  speech  and 
that  they  require  "teratological"*  study  inasmuch  as  they  have 
never  suffered  admixture?  More  important  still,  a  loan-word  no 
longer  counts  as  such  whenever  it  is  studied  within  a  system;  it 
exists  only  through  its  relation  with,  and  opposition  to,  words 
associated  with  it,  just  like  any  other  genuine  sign.  Knowledge  of 
the  circumstances  that  contributed  to  the  development  of  a  lan- 
guage, generally  speaking,  is  never  indispensable.  For  certain 
languages — e.g.  Zend  and  Old  Slavic — even  the  identity  of  the 
original  speakers  is  unknown,  but  lack  of  such  information  in  no 
way  hinders  us  in  studying  these  languages  internally  and  learning 
about  the  transformations  that  they  have  undergone.  In  any  case, 
separation  of  the  two  viewpoints  is  mandatory,  and  the  more 
rigidly  they  are  kept  apart,  the  better  it  will  be. 

The  best  proof  of  the  need  for  separating  the  two  viewpoints  is 
that  each  creates  a  distinct  method.  External  linguistics  can  add 
detail  to  detail  without  being  caught  in  the  vise  of  a  system.  Each 
writer,  for  instance,  will  group  as  he  sees  fit  facts  about  the  spread- 
ing of  a  language  beyond  its  territory.  If  he  looks  for  the  forces 
that  created  a  literary  language  beside  local  dialects,  he  can  always 
use  simple  enumeration.  If  he  arranges  the  facts  more  or  less 
systematically,  he  will  do  this  solely  for  the  sake  of  clarity. 

In  internal  linguistics  the  picture  differs  completely.  Just  any 
arrangement  will  not  do.  Language  is  a  system  that  has  its  own 
arrangement.  Comparison  with  chess  will  bring  out  the  point.  In 
chess,  what  is  external  can  be  separated  relatively  easily  from  what 
is  internal.  The  fact  that  the  game  passed  from  Persia  to  Europe 
is  external ;  against  that,  everything  having  to  do  with  its  system 
and  rules  is  internal.  If  I  use  ivory  chessmen  instead  of  wooden 
ones,  the  change  has  no  effect  on  the  system,  but  if  I  decrease  or 

•  'Pertaining  to  the  study  of  monsters,'  see  p.  54,  footnote.  [Tr.] 


increase  the  number  of  chessmen,  this  change  has  a  profound  effect 
on  the  "grammar"  of  the  game.  One  must  always  distinguish  be- 
tween what  is  internal  and  what  is  external.  In  each  instance  one 
can  determine  the  nature  of  the  phenomenon  by  applying  this 
rule:  everything  that  changes  the  system  in  any  way  is  internal. 

Chapter  VI 

1.  Need  for  Studying  the  Subject 

The  concrete  object  of  linguistic  science  is  the  social  product 
deposited  in  the  brain  of  each  individual,  i.e.  language.  But  the 
product  differs  with  linguistic  groups:  we  have  to  work  with  lan- 
guages. The  linguist  is  obliged  to  acquaint  himself  with  the  greatest 
possible  number  of  languages  in  order  to  determine  what  is  uni- 
versal in  them  by  observing  and  comparing  them. 

But  we  generally  learn  about  languages  only  through  writing. 
Even  in  studying  our  native  language,  we  constantly  make  use  of 
written  texts.  The  necessity  of  using  written  evidence  increases 
when  dealing  with  remote  idioms,  and  all  the  more  when  studying 
idioms  that  no  longer  exist.  We  would  have  direct  texts  at  our  dis- 
posal in  every  instance  only  if  people  had  always  done  what  is  now 
being  done  in  Paris  and  Vienna.  There,  samples  of  all  languages 
are  being  recorded.  Even  so,  recorded  specimens  could  be  made 
available  to  others  only  through  writing. 

Writing,  though  unrelated  to  its  inner  system,  is  used  continually 
to  represent  language.  We  cannot  simply  disregard  it.  We  must  be 
acquainted  with  its  usefulness,  shortcomings,  and  dangers. 

2.  Influence  of  Writing;  Reasons  for  Its  Ascendance 
over  the  Spoken  Form 

Language  and  writing  are  two  distinct  systems  of  signs;  the 
second  exists  for  the  sole  purpose  of  representing  the  first.  The 
linguistic  object  is  not  both  the  written  and  the  spoken  forms  of 


words;  the  spoken  forms  alone  constitute  the  object.  But  the 
spoken  word  is  so  intimately  bound  to  its  written  image  that  the 
latter  manages  to  usurp  the  main  role.  People  attach  even  more 
importance  to  the  written  image  of  a  vocal  sign  than  to  the  sign 
itself.  A  similar  mistake  would  be  in  thinking  that  more  can  be 
learned  about  someone  by  looking  at  his  photograph  than  by 
viewing  him  directly. 

This  illusion,  which  has  always  existed,  is  reflected  in  many  of 
the  notions  that  are  currently  bandied  about  on  the  subject  of 
language.  Take  the  notion  that  an  idiom  changes  more  rapidly 
when  writing  does  not  exist.  Nothing  could  be  further  from  the 
truth.  Writing  may  retard  the  process  of  change  under  certain 
conditions,  but  its  absence  in  no  way  jeopardizes  the  preservation 
of  language.  The  oldest  written  texts  of  Lithuanian,  which  is  still 
spoken  in  eastern  Prussia  and  in  a  part  of  Russia,  date  from  1540; 
but  the  language  of  even  that  late  period  offers  a  more  faithful 
picture  of  Proto-Indo-European  than  does  Latin  of  300  B.C.  This 
one  example  is  enough  to  show  the  extent  to  which  languages  are 
independent  of  writing. 

Certain  very  slight  linguistic  facts  have  been  preserved  without 
the  help  of  any  notation.  During  the  whole  Old  High  German 
period,  people  wrote  tdten,fuolen,  stozen;  near  the  end  of  the  twelfth 
century  the  forms  toten,  f Helen  appeared,  but  stozen  subsisted.  How 
did  the  difference  originate?  Wherever  the  umlaut  occurred,  there 
was  a  ?/  in  the  following  syllable.  Proto-Germanic  had  *daupyan, 
*folyan,  but  *stautan.  At  the  very  beginning  of  the  literary  period 
(about  800)  the  y  became  so  weak  that  no  trace  of  it  appears  in 
writing  for  three  centuries ;  still,  a  slight  trace  had  remained  in  the 
spoken  form ;  that  is  how  it  miraculously  reappeared  as  an  umlaut 
around  1180!  Without  the  help  of  writing,  a  slight  difference  in 
pronunciation  was  accurately  transmitted. 

Thus  language  does  have  a  definite  and  stable  oral  tradition  that 
is  independent  of  writing,  but  the  influence  of  the  written  form 
prevents  our  seeing  this.  The  first  linguists  confused  language  and 
writing,  just  as  the  humanists  had  done  before  them.  Even  Bopp 
failed  to  distinguish  clearly  between  letters  and  sounds.  His  works 
give  the  impression  that  a  language  and  its  alphabet  are  insepa- 


rable.  His  immediate  successors  fell  into  the  same  trap;  the  tran- 
scription th  (for  the  fricative  J^)  caused  Grimm  to  think  not  only 
that  th  was  a  double  sound  but  also  that  it  was  an  aspirated  occlu- 
sive, and  he  accordingly  assigned  it  a  specific  place  in  his  law  of 
consonantal  mutation  or  Lautverschiebung  (see  p.  144).  Scholars 
still  confuse  language  and  writing.  Gaston  Deschamps  said  that 
Berthelot  "had  saved  French  from  ruin"  because  he  had  opposed 
spelling  reform! 

But  how  is  the  influence  of  writing  to  be  explained? 

1)  First,  the  graphic  form  of  words  strikes  us  as  being  something 
permanent  and  stable,  better  suited  than  sound  to  account  for  the 
unity  of  language  throughout  time.  Though  it  creates  a  purely 
fictitious  unity,  the  superficial  bond  of  writing  is  much  easier  to 
grasp  than  the  only  true  bond,  the  bond  of  sound. 

2)  Most  people  pay  more  attention  to  visual  impressions  simply 
because  these  are  sharper  and  more  lasting  than  aural  impressions; 
that  is  why  they  show  a  preference  for  the  former.  The  graphic 
form  manages  to  force  itself  upon  them  at  the  expense  of  sound. 

3)  The  literary  language  adds  to  the  undeserved  importance  of 
writing.  It  has  its  dictionaries  and  grammars;  in  school,  children 
are  taught  from  and  by  means  of  books;  language  is  apparently 
governed  by  a  code ;  the  code  itself  consists  of  a  written  set  of  strict 
rules  of  usage,  orthography ;  and  that  is  why  writing  acquires  pri- 
mary importance.  The  result  is  that  people  forget  that  they  learn 
to  speak  before  they  learn  to  write,  and  the  natural  sequence  is 

4)  Finally,  when  there  is  a  disagreement  between  language  and 
orthography,  settlement  of  the  dispute  is  difiicult  for  everyone 
except  the  linguist;  and  since  he  is  given  no  voice  in  the  matter, 
the  written  form  almost  inevitably  wins  out,  for  any  solution 
supported  by  it  is  easier;  thus  writing  assumes  undeserved  im- 

3.  Systems  of  Writing 

There  are  only  two  systems  of  writing: 

1)  In  an  ideographic  system  each  word  is  represented  by  a  single 
sign  that  is  unrelated  to  the  sounds  of  the  word  itself.  Each  written 



sign  stands  for  a  whole  word  and,  consequently,  for  the  idea  ex- 
pressed by  the  word.  The  classic  example  of  an  ideographic  system 
of  writing  is  Chinese. 

2)  The  system  commonly  known  as  "phonetic"  tries  to  repro- 
duce the  succession  of  sounds  that  make  up  a  word.  Phonetic 
systems  are  sometimes  syllabic,  sometimes  alphabetic,  i.e.,  based 
on  the  irreducible  elements  used  in  speaking. 

Moreover,  ideographic  systems  freely  become  mixtures  when 
certain  ideograms  lose  their  original  value  and  become  symbols  of 
isolated  sounds. 

The  statement  that  the  written  word  tends  to  replace  the  spoken 
one  in  our  minds  is  true  of  both  systems  of  writing,  but  the  tend- 
ency is  stronger  in  the  ideographic  system.  To  a  Chinese,  an 
ideogram  and  a  spoken  word  are  both  symbols  of  an  idea ;  to  him 
writing  is  a  second  language,  and  if  two  words  that  have  the  same 
sound  are  used  in  conversation,  he  may  resort  to  writing  in  order 
to  express  his  thought.  But  in  Chinese  the  mental  substitution  of 
the  written  word  for  the  spoken  word  does  not  have  the  annoying 
consequences  that  it  has  in  a  phonetic  system,  for  the  substitution 
is  absolute;  the  same  graphic  symbol  can  stand  for  words  from 
different  Chinese  dialects. 

I  shall  hmit  discussion  to  the  phonetic  system,  and  especially  to 
the  one  used  today,  the  system  that  stems  from  the  Greek 

'  The  correspondence  between  Saussure'e  system  of  transcription  and  that 
recommended  by  the  International  Phonetic  Association  is  roughly  as  follows : 
































French  pur 


















French  tin 


















French  bon 






French  seul 






French  creuse 






French  un 


When  first  devised  a  phonetic  alphabet — unless  borrowed  and 
already  marked  by  inconsistencies — gives  a  fairly  rational  repre- 
sentation of  language.  With  respect  to  logic,  Greek  is  especially 
noteworthy  (see  p.  64).  But  the  harmonious  relation  between 
writing  and  pronunciation  does  not  last.  Why?  This  question 
must  be  examined. 

4.  Reasons  for  the  Discrepancy  between  Writing  and  Pronunciation 
Of  the  numerous  causes  of  lack  of  agreement  between  writing 
and  pronunciation,  I  shall  recall  only  the  more  important  ones. 
First,  language  is  constantly  evolving,  whereas  writing  tends  to 
remain  stable.  The  result  is  that  a  point  is  reached  where  writing 
no  longer  corresponds  to  what  it  is  supposed  to  record.  A  tran- 
scription that  is  accurate  at  a  particular  moment  will  be  absurd  a 
century  later.  For  a  time  people  may  change  their  graphic  symbols 
to  conform  with  changes  in  pronunciation,  then  relinquish  the 
effort.  This  happened  in  French  in  the  case  of  oi: 

Pronunciation  Written  Forms 

Eleventh  Century 1  rei,  lei  rei,  lei 

Thirteenth  Century  ....  2  roi,  loi  roi,  loi 

Fourteenth  Century  ....  3  roe,  loe  roi,  loi 

Nineteenth  Century  ....  4  rwa,  Iwa  roi,  loi 

Up  until  period  2  changes  in  pronunciation  were  recorded; 
each  step  in  the  history  of  the  language  was  matched  by  a  cor- 
responding step  in  the  history  of  writing.  But  after  the  fourteenth 
century  the  written  form  of  the  words  remained  unchanged  while 
the  evolution  of  the  language  continued;  from  that  moment  the^ 
discrepancy  between  the  language  and  its  orthography  increased 
progressively.  Finally,  the  practice  of  joining  discordant  terms  had 
its  repercussion  on  the  graphic  system  itself:  the  combination  oi 
acquired  a  value  that  was  unrelated  to  either  o  or  i. 












French  blanc 









German  ich 






German  doch 




See  especially  pages  46-49.  [Tr.] 


Such  examples  could  be  multiplied  indefinitely.  For  instance, 
why  should  the  French  write  mats  'but'  and  fait  'fact'  when  the 
words  are  pronounced  me  and  fef  Why  does  c  often  have  the  value 
of  sf  The  answer  is  that  French  has  retained  outmoded  spellings, 

SpeUing  always  lags  behind  pronunciation.  The  I  in  French  is 
today  changing  to  ]j;  speakers  say  eveyer,  mouyer,  just  as  they  say 
essuyer  'wipe,'  nettoyer  'clean';  but  the  written  forms  of  these  words 
are  still  eveiller  'awaken,'  mouiller  'soak.' 

Another  reason  for  discrepancy  between  spelling  and  pronunci- 
ation is  this:  if  an  alphabet  is  borrowed  from  another  language,  its 
resources  may  not  be  appropriate  for  their  new  function;  expedi- 
ents will  have  to  be  found  (e.g.  the  use  of  two  letters  to  designate 
a  single  sound).  Take  the  voiceless  dental  fricative  \>  of  the  Ger- 
manic languages.  Since  Latin  had  no  sign  for  this  sound,  th  was 
used.  The  Merovingian  king  Chilperic  tried  to  add  a  special  symbol 
for  this  sound  to  the  Latin  alphabet,  but  his  attempt  was  unsuc- 
cessful and  ih  won  acceptance.  During  the  Middle  Ages  English 
had  a  closed  e  (e.g.  sed)  and  an  open  e  (e.g.  led) ;  since  the  alphabet 
failed  to  provide  distinct  symbols  for  the  two  sounds,  the  spellings 
seed  and  lead  were  devised.  French  uses  the  double  symbol  ch  to 
stand  for  hushing  I,  etc. 

The  influence  of  etymology  also  helps  to  widen  the  gap  between 
spelling  and  pronunciation.  It  has  been  especially  strong  during 
certain  periods  (e.g.  the  Renaissance).  Even  a  false  etymology 
often  forces  itself  into  the  spelling  of  a  word:  d  was  inserted  in 
French  jpoids  'weight'  as  if  the  word  were  derived  from  Latin 
pondus;  poids  actually  comes  from  pensum.^  Whether  the  appli- 
cation of  the  principle  is  correct  matters  little;  the  fallacy  is  in 
spelling  words  according  to  their  etymology. 

Other  reasons  for  the  discrepancy  are  not  so  obvious;  some 
absurdities  cannot  be  excused  even  on  etymological  grounds.  Why 
was  thun  used  instead  of  tun  in  German?  The  h  was  said  to  repre- 
sent the  aspiration  that  followed  the  initial  consonant ;  but  it  would 
have  to  be  inserted  wherever  aspiration  occurs,  and  many  similar 
words  were  never  written  with  h  (Tugend,  Tisch,  etc.). 

8  Cf.  English  island,  derived  from  ig  'island'  and  land  'land'  but  influenced 
by  isle,  and  doubt,  derived  from  Old  French  douter  but  later  changed  to  con- 
form with  Latin  dubitare.  [Tr.] 


5.  Results  of  the  Discrepancy 

To  classify  the  inconsistencies  of  writing  would  take  too  long. 
One  salient  disadvantage  is  the  multiplicity  of  symbols  that  stand 
for  the  same  sound.  For  z  French  uses  j,  g,  ge  (joU  'pretty/  geler 
'freeze,'  geai  'jay');  for  z,  both  z  and  s;  for  s,  c,  g  and  t  (nation 
'nation'),  sc  (acquiescer  'acquiesce'),  sg  (acquiesgant  'acquiescent'), 
X  {dix  'ten') ;  and  for  k  it  uses  c,  qu,  k,  ch,  cc,  cqu  (acquerir  'acquire'). 
Conversely,  a  single  symbol  stands  for  several  values :  t  stands  for 
t  or  s,  g  for  g  or  z,  etc.^ 

"Indirect  spellings"  also  merit  our  attention.  There  is  no  double 
consonant  in  Zettel,  Teller,  etc. ;  German  uses  tt,  II,  etc.  for  the  sole 
purpose  of  indicating  that  the  preceding  vowel  is  open  and  short. 
Through  a  similar  aberration  English  adds  a  final  silent  e  to 
lengthen  the  preceding  vowel:  mad,  made.  The  e,  which  actually 
affects  only  the  preceding  syllable,  creates  a  second  syllable  for 
the  eye. 

These  irrational  spellings  still  stand  for  something  in  language ; 
but  others  have  neither  rime  nor  reason.  French  has  no  double 
consonants  except  the  old  futures  mourrai  '(I)  shall  die,'  courrai 
*(I)  shall  run,'  etc.;  yet  illegitimate  double  consonants  abound  in 
the  orthography  of  the  language  (bourru  'surly,'  sottise  'foolish- 
ness,' souffrir  'suffer,'  etc.). 

Being  unstable  and  striving  always  for  regularity,  writing  may 
vacillate  at  times ;  the  result  is  fluctuating  orthographies  that  stem 
from  efforts  to  record  sounds  at  different  periods.  Take  ertha,  erdha, 
erda,  or  thrl,  dhri,  dri  in  Old  High  German:  th,  dh,  d  stand  for  the 
same  phonic  element.  But  which  element?  Writing  does  not  provide 
the  answer.  The  complication  that  arises  is  this:  confronted  with 
two  spellings  for  the  same  word,  we  cannot  always  decide  whether 
two  pronunciations  are  actually  represented.  Suppose  that  texts  of 
neighboring  dialects  show  the  spelling  asca  for  a  word  in  one  of  the 
dialects  and  ascha  for  the  same  word  in  the  other;  if  the  sound  is 
the  same,  the  transcriptions  point  to  an  orthographic  fluctuation ; 
if  not,  the  difference  is  phonological  and  dialectal,  as  in  the  Greek 
forms  paizo,  paizdo,  palddo.  Or  two  successive  periods  may  be 

'  The  discrepancy  between  spelling  and  pronunciation  is  of  course  more 
Btriking  in  English  than  in  French:  two  perfectly  riming  sounds  are  WTitten 
fight  and  bite;  c  stands  for  the  same  sound  as  both  s  and  k;  etc.  [Tr.] 


involved.  The  English  forms  hwat,  hweel,  etc.  were  later  replaced 
by  what,  wheel,  etc.  Does  this  point  to  a  graphic  change  or  to  a 
phonetic  change? 

The  preceding  discussion  boils  down  to  this:  writing  obscures 
language ;  it  is  not  a  guise  for  language  but  a  disguise.  That  fact  is 
clearly  illustrated  by  the  spelling  of  French  oiseau  'bird.'  Not  one 
spoken  sound  (wazo)  is  indicated  by  its  own  symbol.  Here  writing 
fails  to  record  any  part  of  the  picture  of  language. 

Another  result  is  that  the  less  writing  represents  what  it  is 
supposed  to  represent,  the  stronger  the  tendency  to  use  it  as  a 
basis  becomes.  Grammarians  never  fail  to  draw  attention  to  the 
written  form.  Psychologically,  the  tendency  is  easily  explained, 
but  its  consequences  are  annoying.  Free  use  of  the  words  "pro- 
nounce" and  "pronunciation"  sanctions  the  abuse  and  reverses 
the  real,  legitimate  relationship  between  writing  and  language. 
Whoever  says  that  a  certain  letter  must  be  pronounced  a  certain 
way  is  mistaking  the  written  image  of  a  sound  for  the  sound  itself. 
For  French  oi  to  be  pronounced  wa,  this  spelling  would  have  to 
exist  independently;  actually  wa  is  written  oi.  To  attribute  the 
oddity  to  an  exceptional  pronunciation  of  o  and  i  is  also  misleading, 
for  this  impUes  that  language  depends  on  its  written  form  and  that 
certain  hberties  may  be  taken  in  writing,  as  if  the  graphic  symbols 
were  the  norm. 

False  notions  about  the  relationship  between  sound  and  graphic 
symbols  appear  even  in  grammatical  rules,  as  in  the  case  of  French 
h.  Some  words  that  begin  with  an  unaspirated  vowel  are  written 
with  h  through  remembrance  of  their  Latin  forms:  homme  'man' 
(formerly  ome)  because  of  Latin  homo.  But  in  words  of  Germanic 
origin,  initial  h  was  actually  pronounced:  hache  'hatchet,'  hareng 
'herring,'  honte  'shame,'  etc.  As  long  as  aspiration  was  used,  words 
of  Germanic  origin  obeyed  the  laws  governing  initial  consonants: 
speakers  said  deu  haches  'two  hatchets,'  le  hereng  'the  herring'; 
other  words  obeyed  the  laws  governing  initial  vowels ;  speakers 
said  deu-z-ommes  'two  men,'  Vomme  'the  man.'  For  that  period  the 
rule,  "Liaison  and  elision  do  not  occur  before  aspirated  /i,"  was 
correct.  But  nowadays  the  formula  is  meaningless.  Aspirated  h  no 
longer  exists  unless  the  label  is  applied  to  something  which  is  not 


a  sound  but  which  prevents  liaison  and  elision.  Again  we  are 
involved  in  a  vicious  circle,  and  h  is  but  a  fictitious  offspring  of 

The  pronunciation  of  a  word  is  determined,  not  by  its  spelling, 
but  by  its  history.  The  form  of  a  word  at  a  particular  moment 
stands  for  a  moment  in  its  enforced  evolution.  Precise  laws  govern 
its  evolution.  Each  step  is  determined  by  the  preceding  step.  The 
only  thing  to  consider  is  the  one  most  often  forgotten :  the  evolution 
of  the  word,  its  etymology. 

The  name  of  the  town  of  Auch  is  o§  in  phonetic  transcription. 
That  is  the  only  French  word  in  which  final  ch  stands  for  I.  But  we 
explain  nothing  by  saying,  "Final  ch  is  pronounced  5  only  in  Auch." 
The  only  question  that  concerns  us  is  this :  How  could  Latin  Auscii 
have  changed  to  o§?  Orthography  is  unimportant. 

Should  French  gageure  'wager'  be  pronounced  with  o  or  ii?  Some 
speakers  say :  gazor,  for  heure  'hour'  is  pronounced  dr.  Others  say : 
No,  it  is  gazilr,  for  ge  is  equivalent  z,  as  in  gedle  'jail.'  The  argument 
is  pointless.  The  real  issue  is  etymological :  gageure  was  formed  from 
gager  'earn'  just  as  tournure  'figure'  was  formed  from  tourner  'turn'; 
only  gaziir  is  justifiable;  gazor  is  due  solely  to  the  equivocal  nature 
of  writing. 

But  the  tyranny  of  writing  goes  even  further.  By  imposing  itself 
upon  the  masses,  spelling  influences  and  modifies  language.  This 
happens  only  in  highly  literate  languages  where  written  texts  play 
an  important  role.  Then  visual  images  lead  to  wrong  pronunci- 
ations; such  mistakes  are  really  pathological.^"  Spelling  practices 
cause  mistakes  in  the  pronunciation  of  many  French  words.  For 
instance,  there  were  two  spellings  for  the  surname  Lef^vre  (from 
Latin /a6er),  one  popular  and  simple,  the  other  learned  and  ety- 
mological: Lefevre  and  Lefebvre.  Because  v  and  u  were  not  kept 
apart  in  the  old  system  of  writing,  Lefebvre  was  read  as  Lefebure, 
with  a  b  that  had  never  really  existed  and  a  u  that  was  the  result 
of  ambiguity.  Now,  the  latter  form  is  actually  pronounced. 

Mispronunciations  due  to  spelling  will  probably  appear  more 
frequently  as  time  goes  on,  and  the  number  of  letters  pronounced 

'"  Pathology  was  given  currency  in  French  by  Littr6.  It  was  used  subse- 
quently by  Gilli6ron  and  Darmsteter  as  well  as  by  Saussure.  See  note  6.  [Tr.] 


by  speakers  will  probably  increase.  Some  Parisians  already  pro- 
nounce the  t  in  sept  femmes  'seven  women' ;^^  Darmsteter  foresees 
the  day  when  even  the  last  two  letters  of  vingt  'twenty'  will  be 
pronounced — truly  an  orthographic  monstrosity. 

Such  phonic  deformations  belong  to  language  but  do  not  stem 
from  its  natural  functioning.  They  are  due  to  an  external  influence. 
Linguistics  should  put  them  into  a  special  compartment  for  obser- 
vation: they  are  teratological  cases.^ 

Chapter  VII 

1.  Definition 

Whoever  consciously  deprives  himseK  of  the  perceptible  image 
of  the  written  word  runs  the  risk  of  perceiving  only  a  shapeless  and 
unmanageable  mass.  Taking  away  the  written  form  is  like  depriv- 
ing a  beginning  swimmer  of  his  life  belt. 

To  substitute  immediately  what  is  natural  for  what  is  artificial 
would  be  desirable;  but  this  is  impossible  without  first  studying 
the  sounds  of  language ;  apart  from  their  graphic  symbols,  sounds 
are  only  vague  notions,  and  the  prop  provided  by  writing,  though 
deceptive,  is  still  preferable.  The  first  linguists,  w^ho  knew  nothing 
about  the  physiology  of  articulated  sounds,  were  constantly  falling 
into  a  trap ;  to  me,  it  means  a  first  step  in  the  direction  of  truth,  for 
the  study  of  sounds  themselves  furnishes  the  desired  prop.  Modern 

"  The  pronunciation  [se]  is  now  obsolescent.  Cf.  the  trend  toward  pro- 
nouncing the  t  in  often.  [Tr.] 

'2  Saussure's  terminology  is  reminiscent  of  the  biological  parlance  of  Gillieron 
(e.g.  in  Pathologie  et  therapeidique  verbales,  Paris,  1921).  [Tr.] 

'^  Saussure  later  modifies  and  expands  his  definition  of  phonology  (see 
especially  pp.  34,  42  ff.,  117  ff.  and  131).  Only  M.  Grammont  has  followed 
Saussure's  practice.  English  and  American  linguists  often  use  phonology  to 
indicate  the  historical  study  of  sounds  or  the  study  of  the  functioning  of 
Bounds  in  a  particular  language,  phonetics  for  the  study  of  the  modaUtiea 
of  sounds  used  in  speaking,  and  phonemics  (corresponding  to  French  phonologie 
and  German  Phonologie)  for  the  study  of  the  distinctive  sounds  of  language. 


linguists  have  finally  seen  the  light;  pursuing  for  their  own  ends 
investigations  started  by  others  (physiologists,  theoreticians  of 
singing,  etc.),  they  have  given  linguists  an  auxiliary  science  that 
has  freed  it  from  the  written  word. 

The  physiology  of  sounds  (German  Laut-  or  Sprachphysiologie) 
is  often  called  phonetics  (French  phonetique,  German  Phonetik) .  To 
me  this  name  seems  inappropriate.  Instead,  I  shall  use  -phonology. 
For  phonetics  first  designated — and  should  continue  to  designate — 
the  study  of  the  evolutions  of  sounds.  Two  absolutely  distinct  dis- 
ciplines should  not  be  lumped  together  under  the  same  name. 
Phonetics  is  a  historical  science;  it  analyses  events  and  changes, 
and  moves  through  time.  Phonology  is  outside  time,  for  the  ar- 
ticulatory  mechanism  never  changes. 

The  two  studies  are  distinct  but  not  opposites.  Phonetics  is  a 
basic  part  of  the  science  of  language;  phonology — this  bears 
repeating — is  only  an  auxiliary  discipline  and  belongs  exclusively  to 
speaking  (see  pp.  17  ff.).  Just  what  phonational  movements  could 
accomplish  if  language  did  not  exist  is  not  clear;  but  they  do  not 
constitute  language,  and  even  after  we  have  explained  all  the  move- 
ments of  the  vocal  apparatus  necessary  for  the  production  of  each 
auditory  impression,  we  have  in  no  way  illuminated  the  problem 
of  language.  It  is  a  system  based  on  the  mental  opposition  of  audi- 
tory impressions,  just  as  a  tapestry  is  a  work  of  art  produced  by 
the  visual  oppositions  of  threads  of  different  colors;  the  important 
thing  in  analysis  is  the  role  of  the  oppositions,  not  the  process 
through  which  the  colors  were  obtained. 

An  outUne  of  the  phonological  system  is  given  in  the  Appendix; 
here  I  am  trying  merely  to  determine  the  extent  to  which  pho- 
nology can  help  linguistics  to  escape  the  delusions  of  writing. 

2.  Phonological  Writing 

The  linguist  needs  above  all  else  a  means  of  transcribing  articu- 
lated sounds  that  will  rule  out  all  ambiguity.  Actually,  countless 
graphic  systems  have  been  proposed. 

What  are  the  requirements  for  a  truly  phonological  system  of 
writing?  First,  there  should  be  one  symbol  for  each  element  of  the 
spoken  chain.  This  requirement  is  not  always  considered.  Thus 
English  phonologists,  concerned  with  classification  rather  than 


analysis,  have  two-  and  three-letter  S5anbols  for  certam  sounds. 
Second,  there  should  be  some  means  for  making  a  rigid  distinction 
between  implosive  and  explosive  sounds  (see  pp.  49  ff.). 

Are  there  grounds  for  substituting  a  phonological  alphabet  for 
a  system  already  in  use?  Here  I  can  only  broach  this  interesting 
subject.  I  think  that  phonological  writing  should  be  for  the  use  of 
linguists  only.  First,  how  would  it  be  possible  to  make  the  English, 
Germans,  French,  etc.  adopt  a  uniform  system!  Next,  an  alphabet 
applicable  to  all  languages  would  probably  be  weighed  down  by 
diacritical  marks;  and — to  say  nothing  of  the  distressing  appear- 
ance of  a  page  of  phonological  writing — attempts  to  gain  precision 
would  obviously  confuse  the  reader  by  obscuring  what  the  writing 
was  designed  to  express.  The  advantages  would  not  be  sufficient 
to  compensate  for  the  inconveniences.  Phonological  exactitude  is 
not  very  desirable  outside  science. 

Reading  is  another  issue.  We  read  in  two  ways:  a  new  or  un- 
known word  is  spelled  out  letter  by  letter;  but  a  common,  ordinary 
word  is  embraced  by  a  single  glance,  independently  of  its  letters, 
so  that  the  image  of  the  whole  word  acquires  an  ideographic  value. 
Here  traditional  orthography  takes  revenge.  It  is  useful  to  dis- 
tinguish between  French  tant  'so  much'  and  temps  'weather'; 
et  'and,'  est  'is,'  and  ait  'have';  du  'of  the'  and  diX  'had  to';  il  devait 
*he  owed'  and  Us  devaienl  'they  owed,'  etc.^*  Let  us  hope  only  that 
the  most  flagrant  absurdities  of  writing  will  be  eliminated.  Al- 
though a  phonological  alphabet  is  helpful  in  the  teaching  of  lan- 
guages, its  use  should  not  be  generalized. 

3.  Validity  of  Evidence  Furnished  by  Writing 

One  must  not  think  that  spelling  reform  should  immediately 
follow  the  realization  that  writing  is  deceptive.  The  genuine  con- 
tribution of  phonology  is  in  providing  precautionary  measures  for 
dealing  with  the  written  form  through  which  we  must  pass  in  order 
to  reach  language.  Evidence  furnished  by  writing  is  valid  only 
when  interpreted.  We  must  draw  up  for  each  language  studied  a 
phonological  system,  i.e.  a  description  of  the  sounds  with  which  it 
functions;  for  each  language  operates  on  a  fixed  number  of  well- 
differentiated  phonemes.  This  system  is  the  only  set  of  facts  that 
;    "  Cf.  English  sow  and  sew;  to,  too,  and  two;  due  and  dew,  etc.  [Tr.] 


interests  the  linguist.  Graphic  symbols  bear  but  a  faint  resem- 
blance to  it;  the  difficulty  of  determining  the  accuracy  of  the 
resemblance  varies  according  to  the  idiom  and  circumstances. 

The  linguist  who  deals  with  a  language  of  the  past  has  only  in- 
direct data  at  his  disposal.  What  resources  can  he  use  in  setting 
up  its  phonological  system? 

1)  First  and  foremost  is  external  evidence,  especially  contem- 
porary descriptions  of  the  sounds  and  pronunciations  of  the  period. 
French  grammarians  of  the  sixteenth  and  seventeenth  centuries, 
especially  those  interested  in  teaching  foreigners,  left  us  many 
interesting  observations.  But  the  information  contained  in  the 
writings  of  contemporaries  is  often  vague,  for  the  authors  have  no 
phonological  method.  The  terminology  of  their  descriptions  is 
whimsical  and  lacks  scientific  precision.  The  result  is  that  their 
evidence  must  in  turn  be  interpreted.  Names  given  to  sounds,  for 
instance,  are  often  misleading:  Greek  grammarians  called  voiced 
b,  d,  g,  etc.  "middle"  consonants  (mesai),  and  voiceless  p,  t,  k,  etc. 
psllai,  which  Latin  grammarians  translated  by  tenues. 

2)  More  accurate  information  will  result  from  combining  ex- 
ternal data  with  internal  evidence,  which  I  shall  class  under  two 

a)  The  first  class  comprises  evidence  based  on  the  regidarity  of 
phonetic  evolutions.  Knowing  what  sound  a  letter  stood  for  during 
another  period  is  important  in  determining  the  value  of  that  letter. 
Its  present  value  is  the  result  of  an  evolution  that  allows  us  to  cast 
aside  certain  hypotheses  from  the  outset.  For  instance,  the  exact 
value  of  Sanskrit  q  is  unknown,  but  the  fact  that  it  replaced  palatal 
Proto-Indo-European  k  clearly  limits  the  field  of  conjecture. 

If  the  hnguist  knows  both  the  point  of  departure  and  the  parallel 
evolution  of  similar  sounds  of  a  particular  language  during  the 
same  period,  he  can  use  analogical  reasoning  and  set  up  a  pro- 

Naturally,  the  problem  of  determining  an  intermediate  pro- 
nunciation is  easier  when  both  the  starting  point  and  the  end 
result  are  unknown.  French  an  (e.g.  in  sauter  'jump')  must  have 
been  a  diphthong  during  the  Middle  Ages,  for  it  is  half-way  be- 
tween older  al  and  modern  o.  And  if  we  learn  by  some  other 
means  that  the  diphthong  still  existed  at  a  particular  moment,  we 


are  safe  in  assuming  that  it  also  existed  during  the  preceding  period. 
We  do  not  know  exactly  what  z  stands  for  in  a  word  like  Old  High 
German  wazer;  but  our  guideposts  are  the  older  form  water  on  the 
one  hand  and  Modern  German  Wasser  on  the  other.  The  z  must  be 
a  sound  half-way  between  t  and  s;  we  can  reject  any  hypothesis 
that  fails  to  consider  both  t  and  s;  to  hold  that  z  stands  for  a  palatal 
sound,  for  example,  would  be  impossible,  for  only  a  dental  articu- 
lation can  logically  come  between  two  other  dental  articulations. 

b)  There  are  several  types  of  contemporary  evidence.  Spelling 
differences  furnish  one  of  many  types.  During  one  period  we  find 
that  Old  High  German  has  wazer,  zehan,  ezan  but  never  wacer, 
cehan,  etc.  When  we  find  the  forms  esan  and  essan,  waser  and 
wasser,  etc.,  however,  we  easily  conclude  that  the  sound  of  z  was 
close  to  s  but  different  from  the  sound  that  c  stood  for  during  the 
same  period.  The  subsequent  appearance  of  such  forms  as  wacer 
proves  that  the  two  originally  distinct  phonemes  became  somewhat 

Poetic  texts  are  invaluable  documents  in  the  study  of  pro- 
nunciation. They  furnish  many  types  of  information,  depending  on 
whether  the  system  of  versification  is  based  on  the  number  of  syl- 
lables, quantity,  or  similarity  of  sounds  (alliteration,  assonance, 
and  rime).  Greek  indicated  certain  long  vowels  in  writing  (e.g. 
o,  transcribed  co)  but  not  others.  We  must  consult  the  poets  in 
order  to  find  out  about  the  quantity  of  a,  i,  and  u.  Thus  rime  allows 
us  to  determine  until  what  period  the  final  consonants  of  Old 
French  gras  and  faz  (Latin  facio  '1  do')  were  different  and  from 
what  moment  they  were  brought  together  and  merged.  Rime  and 
assonance  also  show  that  e  derived  from  Latin  a  (e.g.  yere  'father* 
from  patrem,  tel  'such'  from  talem,  mer  'sea'  from  mare)  was  not 
pronounced  like  other  e's.  These  words  never  appear  in  rime  or 
assonance  with  elle  'she'  (from  ilia),  vert  'green'  (from  viridem), 
belle  'beautiful'  (from  bella),  etc. 

Finally  there  is  the  evidence  furnished  by  the  spelling  of  loan- 
words, puns,  cock-and-bull  stories,  etc.  In  Gothic,  for  example, 
kawtsjo  reveals  information  about  the  pronunciation  of  cautio  in 
Vulgar  Latin.  That  French  roi  'king'  was  pronounced  rwe  at  the 
end  of  the  eighteenth  century  is  attested  by  the  following  story 
cited  by  Nyrop   (Grammaire  historique  de  la  langue  frangaise. 


p.  178) :  A  woman  who  had  been  brought  before  the  revolutionary 
tribunal  was  asked  whether  she  had  not  said  in  the  presence  of 
witnesses  that  a  king  {roi)  was  needed;  she  replied  "that  she  was 
not  speaking  of  a  king  like  Capet  or  the  others  at  all,  but  of  a 
rouet  mattre  'spinning  wheel.'  " 

All  the  foregoing  procedures  help  us  to  acquire  some  knowledge 
of  the  phonological  system  of  a  period  as  well  as  to  interpret  and 
use  profitably  the  evidence  furnished  by  writing. 

In  dealing  with  a  living  language,  the  only  rational  method 
consists  of  (a)  setting  up  the  system  of  sounds  as  revealed  by  direct 
observation,  and  (b)  observing  the  system  of  signs  used  to  repre- 
sent— imperfectly — these  sounds.  Many  grammarians  still  hold 
to  the  old  method  that  I  have  criticized  and  simply  tell  how  each 
letter  is  pronounced  in  the  language  they  wish  to  describe.  By  using 
the  older  method,  however,  they  cannot  present  clearly  the  pho- 
nological system  of  an  idiom. 

Nevertheless,  great  strides  in  the  right  direction  have  already 
been  taken,  and  phonologists  have  made  an  important  contribution 
toward  reforming  our  ideas  about  writing  and  spelling. 

Principles  of  Phonology 

Chapter  I 

L  Definition  of  the  Phoneme 

[For  this  part  we  were  able  to  use  a  stenographic  reproduction  of 
three  lectures  given  by  Saussure  in  1897,  "Theorie  de  la  syllabe," 
in  which  he  also  touches  upon  the  general  principles  discussed  in 
Chapter  I;  moreover,  much  of  the  material  in  his  personal  notes 
deals  with  phonology;  on  many  points,  the  notes  illuminate  and 
complete  the  data  furnished  by  Courses  I  and  III.  (Editors'  note.)] 

Many  phonologists  limit  themselves  almost  exclusively  to  the 
phonational  act,  i.e.  the  production  of  sound  by  the  vocal  organs 
(larynx,  mouth,  etc.)  and  neglect  the  auditory  side.  Their  method 
is  wrong.  Not  only  does  the  auditory  impression  come  to  us  just 
as  directly  as  the  image  of  the  moving  vocal  organs,  but  it  is  also 
the  basis  of  any  theory.  Auditory  impressions  exist  unconsciously 
before  phonological  units  are  studied;  our  ear  tells  us  what  b,  t,  etc. 
are.  Even  if  all  the  movements  made  by  the  mouth  and  larynx  in 
pronouncing  a  chain  of  sounds  could  be  photographed,  the  ob- 
server would  still  be  unable  to  single  out  the  subdivisions  in  the 
series  of  articulatory  movements;  he  would  not  know  where  one 
sound  began  and  the  next  one  ended.  Without  the  auditory  im- 
pression, how  can  we  say  that  in  fal,  for  instance,  there  are  three 
units  rather  than  two  or  four?  But  when  we  hear  a  sound  in  a 
spoken  chain,  we  can  identify  it  immediately;  as  long  as  there  is 
an  impression  of  homogeneity,  the  sound  is  unique.  What  matters 
is  not  the  length  of  the  sound  (cf .  fdl  and  fdl)  but  the  quahty  of  the 
impression.  The  sound-chain  is  not  divided  into  equal  beats  but 
into  homogeneous  ones;  each  beat  is  characterized  by  unity  of 
impression,  and  that  is  the  natural  point  of  departure  for 



Here  the  early  Greek  alphabet  is  noteworthy.  Each  simple 
sound  is  represented  in  Greek  by  a  single  graphic  sign,  and  each 
sign  always  stands  for  the  same  simple  sound.  The  Greek  alphabet 
was  an  ingenious  discovery  that  was  later  handed  down  to  the 
Romans.  In  the  transcription  of  bdrharos  'barbarian/  each  letter 
corresponds  to  a  homogeneous  beat: 


In  the  drawing  above,  the  horizontal  line  stands  for  the  phonetic 
chain,  and  the  short  vertical  bars  indicate  passage  from  one  sound 
to  another.  In  the  early  Greek  alphabet  there  are  no  complex 
graphs  like  English  sh  for  s,  no  interchangeable  letters  for  single 
sounds  like  c  and  s  for  s,  no  single  signs  for  double  sounds  like  x  for 
ks.  A  one-to-one  ratio  between  sounds  and  graphs — the  necessary 
and  sufficient  basis  for  a  good  phonological  system  of  writing — was 
realized  almost  completely  by  the  Greeks.^ 

Other  nations  did  not  grasp  this  principle,  and  their  alphabets 
do  not  analyze  the  spoken  chain  according  to  its  homogeneous 
auditory  beats.  The  Cypriots,  for  example,  stopped  at  more  com- 
plex units  like  pa,  ti,  do,  etc.  Such  notation  is  called  syllabic,  but 
this  name  is  hardly  accurate  since  there  are  still  other  types  of 
syllables  (e.g.  pak,  tra,  etc.).  The  Semites  indicated  only  the  con- 
sonants. They  would  have  transcribed  a  word  like  bdrbaros  as 

Delimitation  of  the  sounds  of  the  spoken  chain  can  be  based  only 
on  auditory  impressions;  but  description  of  these  sounds  is  an 
entirely  different  process.  Description  can  be  carried  out  only  on 

1  To  be  sure,  they  wrote  X,  0,  0  for  kh,  th,  ph;  <^EPO  stands  for  ph^ro; 
but  this  is  a  later  innovation;  archaic  inscriptions  read  KHAPIS  and  not 
XAPIS.  The  same  inscriptions  have  two  signs  for  k,  kappa  and  koppa,  but 
the  situation  is  different:  two  real  differences  in  pronunciation  were  involved, 
A;  being  sometimes  palatal  and  sometimes  velar;  besides,  koppa  later  dis- 
appeared. Finally — and  this  is  a  more  subtle  point — in  early  Greek  and  Latin 
inscriptions  a  double  consonant  is  often  indicated  by  a  simple  letter  (e.g. 
Latin  fuisse,  written  FUISE) ;  this  is  an  infraction  of  the  principle  since  the 
doul)le  s  lasts  two  beats — beats  that  are  not  homogeneous,  as  we  shall  see 
later,  and  that  make  distinct  impressions;  but  the  mistake  is  excusable  since 
the  two  sounds  have  a  common  characteristic  even  though  they  are  distinct 
(cf.  pp.  51  ff.).  [S.] 


the  basis  of  the  articulatory  act,  for  it  is  impossible  to  analyze  the 
units  of  sound  in  their  own  chain.  We  must  go  back  to  the  move- 
ments involved  in  phonation ;  there,  a  given  sound  obviously  cor- 
responds to  a  given  act:  b  (auditory  beat)  =  b'  (articulatory  beat). 
The  first  units  obtained  by  cutting  the  spoken  chain  are  made  up 
of  b  and  b' ;  they  are  phonemes;  a  phoneme  is  the  sum  of  the  audi- 
tory impressions  and  articulatory  movements,  the  unit  heard  and 
the  unit  spoken,  each  conditioning  the  other:  thus  it  is  a  complex 
unit  with  a  foot  in  each  chain. 

The  elements  first  obtained  through  analysis  of  the  spoken  chain 
are  like  the  links  of  this  chain :  they  are  irreducible  moments  that 
cannot  be  studied  outside  the  time  that  they  occupy.  A  grouping 
Uke  ta,  for  instance,  will  always  be  one  moment  plus  another,  one 
fragment  of  a  certain  length  plus  another.  Against  this,  the  ir- 
reducible t,  taken  separately,  can  be  studied  in  the  abstract,  outside 
time.  We  can  speak  of  t  in  general  as  the  T  species  (I  use  capitals 
to  indicate  species),  of  i  in  general  as  the  /  species,  etc.  if  we  con- 
sider only  the  distinctive  character  of  a  sound  and  neglect  every- 
thing that  depends  on  succession  in  time.  Similarly,  a  musical 
series  do,  re,  mi  can  be  treated  only  as  a  concrete  series  in  time, 
but  if  I  select  one  of  its  irreducible  elements,  I  can  study  it  in  the 

Having  analyzed  a  sufficient  number  of  spoken  chains  from 
different  languages,  the  phonologist  can  identify  and  classify  the 
elements  with  which  each  language  operates.  Then,  if  he  ignores 
acoustically  unimportant  variations,  he  will  find  that  the  number 
of  species  is  not  indefinite.  Special  works  hst  these  species  and 
describe  them  in  detail.^  Here  I  wish  merely  to  show  the  simple, 
invariable  principles  upon  which  any  such  classification  is  based. 

But  first  let  me  say  a  few  words  about  the  vocal  apparatus,  the 
possible  functioning  of  the  different  organs,  and  the  role  of  these 
same  organs  as  producers  of  sound. 

2  Cf .  Sievers,  Grundziige  der  Phonetik,  fifth  ed.,  1902;  Jespersen,  Lehrbtich 
der  Phonetik,  second  ed.,  1913;  Roudet,  Elements  de  phonetique  generale, 
1910.  [Ed.] 



2.  The  Vocal  Apparatus  and  Its  Functioning^ 

1)  I  limit  description  of  the  vocal  apparatus  to  a  schematic 
drawing  in  which  A  designates  the  nasal  cavity,  B  the  oral  cavity, 
and  C  the  larynx  (with  the  glottis  e  between  the  two  vocal  cords) . 

In  the  mouth,  the  parts  of  the  vocal  apparatus  that  should  be 
singled  out  are  these :  the  lips  a  and  a;  the  tongue  jS — y  (/?  designat- 
ing the  point  and  7  the  rest) ;  the  upper  teeth  d;  the  palate,  made 
up  of  the  bony  hard  palate  f-h  in  the  front  and  the  movable  mem- 
brane or  soft  palate  i  in  the  back;  and,  finally,  the  uvula  5. 

The  Greek  letters  indicate  organs  that  are  active  during  articu- 
lation ;  the  Latin  letters  identify  the  passive  parts. 

The  glottis  z,  made  up  of  two  parallel  muscles  or  vocal  cords, 
opens  when  the  cords  are  drawn  apart  and  closes  when  they  come 
together.  Complete  closure  does  not  occur;  the  opening  is  some- 
times wide,  sometimes  narrow.  When  the  opening  is  wide,  allowing 

'  Saussure's  brief  description  has  been  supplemented  by  material  based  on 
Jespersen's  Lehrbuch  der  Phonetik,  from  which  we  have  also  borrowed  the 
principle  used  in  setting  up  the  table  of  phonemes  below  (see  pp.  44  ff.).  But 
we  are  merely  carrying  out  Saussure's  intent,  and  the  reader  may  be  assured 
that  these  additions  do  not  alter  his  thought  in  any  way.  [Ed.] 


the  air  to  pass  freely,  no  vibration  is  heard;  voicing  occurs  when 
air  passes  through  a  narrow  opening,  causing  the  cords  to  vibrate. 
There  is  no  other  alternative  in  the  normal  emission  of  sounds. 

The  nasal  cavity  is  a  completely  immobile  organ ;  the  stream  of 
air  can  be  stopped  only  by  raising  the  uvula  6;  it  is  an  open  or  a 
closed  door. 

The  oral  cavity  offers  a  wide  range  of  possibilities;  the  lips  can 
be  used  to  increase  the  length  of  the  channel,  the  jaws  can  be 
puffed  out  or  drawn  in,  and  a  great  variety  of  movements  of  the 
lips  and  tongue  can  be  used  to  contract  or  even  to  close  the  cavity. 

The  role  played  by  the  same  organs  in  producing  sounds  is 
directly  proportional  to  their  mobility ;  uniformity  in  the  function- 
ing of  the  larynx  and  nasal  cavity  is  matched  by  diversity  in  the 
functioning  of  the  oral  cavity. 

Air  that  is  expelled  from  the  lungs  first  passes  through  the 
glottis.  It  is  possible  to  produce  a  laryngeal  sound  by  tightening 
the  vocal  cords,  but  the  larynx  cannot  produce  phonological 
varieties  that  allow  us  to  separate  and  classify  the  sounds  of  lan- 
guage; in  this  respect,  the  laryngeal  sound  is  uniform.  Perceived 
directly  as  it  emitted  by  the  glottis  the  sound  seems  to  have  an 
almost  invariable  quality. 

The  nasal  channel  serves  as  nothing  more  than  a  resonator  for 
the  vocal  vibrations  that  pass  through  it.  It  does  not  function  as 
a  producer  of  sound. 

The  oral  cavity,  on  the  contrary,  functions  both  as  a  producer 
of  sound  and  as  a  resonator.  When  the  glottis  is  wide-open,  there 
is  no  laryngeal  vibration ;  the  sound  that  is  heard  originates  in  the 
oral  cavity  (I  leave  to  the  physicist  the  task  of  deciding  whether 
it  is  a  sound  or  merely  a  noise).  But  when  tightening  of  the  vocal 
cords  causes  the  glottis  to  vibrate,  the  mouth  serves  mainly  to 
modify  the  laryngeal  sound. 

In  short,  the  factors  involved  in  the  production  of  sound  are 
expiration,  oral  articulation,  vibration  of  the  larynx,  and  nasal 

But  simple  enumeration  does  not  identify  the  differential  prop- 
erties of  phonemes.  In  classifying  phonemes,  what  constitutes  them 
is  of  much  less  importance  than  what  distinguishes  them  from  each 
other.  A  negative  force  can  be  more  important  in  classifying  a 



phoneme  than  a  positive  one.  Thus  expiration,  a  positive  element 
that  is  part  of  every  phonational  act,  has  no  dififerentiating  value; 
but  nasal  resonance  may  characterize  phonemes  by  its  absence,  a 
negative  force,  just  as  well  as  by  its  presence.  The  important  thing 
is  that  two  of  the  elements  enumerated  above  are  constant,  and 
that  they  are  necessary  and  sufficient  for  the  production  of  sound : 

a)  expiration 

6)  oral  articulation; 

whereas  the  other  two  may  be  either  absent  or  superimposed  on 
the  first  two: 

c)  vibration  of  the  larynx 

d)  nasal  resonance. 

Moreover,  we  know  that  while  a,  c,  and  d  are  uniform,  b  makes 
possible  the  production  of  many  varieties  of  sounds. 

We  should  also  bear  in  mind  that  a  phoneme  is  identified  when  its 
phonational  act  is  determined,  and  that  all  species  of  phonemes  will 
be  determined  when  all  phonational  acts  are  identified.  The  fore- 
going classification  of  forces  involved  in  the  production  of  sound 
shows  that  phonational  acts  are  differentiated  only  by  b,  c,  and  d. 
For  each  phoneme  we  must  determine  its  oral  articulation,  whether 
a  laryngeal  sound  is  present  (— -)  or  absent  ([  ]),  and  whether  nasal 
resonance  is  present  (....)  or  absent  ([  ]).  When  one  of  these  three 
is  unknown,  the  identification  of  a  sound  is  incomplete.  But  as  soon 
as  all  three  are  known,  their  different  combinations  determine  all 
the  basic  species  of  phonational  acts. 

The  following  table  gives  the  possible  variations: 





a  Expiration 

6  Oral  Articulation 

c  [     ] 

d  [     ] 


Oral  Articulation 

[     ] 

Oral  Articulation 

[     ] 


Oral  Articulation 

Column  I  designates  voiceless  sounds,  II  voiced  sounds.  III 
voiceless  nasalized  sounds,  and  IV  voiced  nasalized  sounds. 

But  one  unknown  remains:  the  nature  of  the  oral  articulation; 


therefore,  the  most  important  thing  is  to  determine  the  possible 
varieties  of  oral  articulation. 

3.  Classification  of  Sounds  According  to  Their  Oral  Articulation 

Sounds  are  generally  classed  according  to  the  place  of  their 
articulation.  My  point  of  departure  will  be  different.  Regardless 
of  where  articulation  takes  place,  there  is  always  a  certain  aperture, 
i.e.,  a  certain  degree  of  opening  that  ranges  between  two  extremes, 
complete  closure  and  maximum  opening.  On  that  basis,  and  pro- 
ceeding from  minimum  to  maximum  aperture,  sounds  will  fall  into 
seven  categories  that  I  shall  designate  by  the  numbers  0,  1,2,  3,  4, 
5,  6.  Only  within  each  category  shall  I  distribute  phonemes  into 
different  types  according  to  their  place  of  articulation. 

I  shall  conform  to  current  terminology  even  though  it  is  im- 
perfect or  incorrect  at  many  points:  words  like  guttural,  palatal, 
dental,  liquid,  etc.  are  all  more  or  less  illogical.  A  more  rational 
plan  would  be  to  divide  the  palate  into  a  certain  number  of  areas. 
Then  by  focusing  attention  on  lingual  articulation,  it  would  always 
be  possible  to  specify  the  main  point  of  contact.  In  devising  a 
formula,  I  shall  draw  upon  this  notion  and  use  the  letters  of  the 
sketch  of  the  vocal  apparatus  (see  p.  41) :  the  number  of  the  aper- 
ture is  placed  between  a  Greek  letter  (indicating  an  active  organ) 
and  a  Latin  letter  (indicating  a  passive  organ).  Thus  /30e  means 
that  complete  closure  is  maintained  while  the  tip  of  the  tongue  is 
placed  against  the  upper  alveolar  ridge. 

Finally,  within  each  articulation  the  different  species  of  pho- 
nemes are  marked  by  concomitant  features — laryngeal  sound  and 
nasal  resonance — which  differentiate  by  their  absence  as  well  as 
by  their  presence. 

The  two  accompanying  features  and  the  formula  provide  a 
simple,  rational  means  of  classifying  phonemes.  Of  course,  one 
should  not  expect  to  find  here  phonemes  that  have  a  complex  or 
special  character,  regardless  of  their  practical  importance  (e.g.  the 
aspirates  ph,  dh,  etc. ;  the  affricates  t§,  dz,  pf,  etc. ;  palatalized  con- 
sonants; weak  vowels  like  9  or  mute  e,  etc.).  Nor  should  one  expect 
to  find  simple  phonemes  that  have  no  practical  importance  and 
that  are  not  considered  differentiated  sounds. 


A.  Zero  Aperture:  Occlusives 

Occlusives  include  all  phonemes  produced  by  complete  closure, 
the  airtight  but  brief  sealing  of  the  oral  cavity.  This  is  not  the  place 
to  discuss  whether  a  sound  is  produced  when  closure  or  release 
occurs;  actually  it  may  be  produced  in  either  way  (see  pp.  51  ff.). 

The  three  main  types  of  occlusives  are  named  according  to  their 
places  of  articulation:  labials  (p,  h,  m);  dentals  {t,  d,  n);  and 
gutturals  (A;,  g,  n). 

The  first  type  is  articulated  with  the  lips;  for  the  second,  the  tip 
of  the  tongue  is  placed  against  the  front  of  the  palate ;  for  the  third, 
the  back  of  the  tongue  makes  contact  with  the  back  part  of  the 

Many  languages,  notably  the  Indo-European,  make  a  distinc- 
tion between  two  guttural  articulations,  one  palatal  (in  the  f-h 
area)  and  the  other  velar  (in  the  i  area),  but  elsewhere  (e.g.  in 
English)  the  difference  goes  unnoticed  and  the  ear  likens  a  back 
k  (such  as  the  sound  of  c  in  cart)  to  a  front  k  (as  in  king). 

The  following  table  gives  the  formulas  for  the  various  occlusive 
phonemes : 




P           b 






g           (n) 

aOa       aOa 

[     ]       — 
[     ]       [     ] 



[     ] 
[     ] 


[     ] 



[      ] 
[      ] 

70h       TOh 

[     ]        .... 

Nasal  m,  n,  and  n  are  really  voiced  nasalized  occlusives;  in  pro- 
nouncing amba,  one  raises  the  uvula  to  close  the  nasal  fossae  in 
shifting  from  m  to  b. 

In  theory,  each  type  has  a  voiceless  nasal — a  nasal  sound  un- 
accompanied by  glottal  vibration;  thus  voiceless  m  occurs  after  a 
voiceless  sound  in  the  Scandinavian  languages;  French  also  has 
voiceless  nasals,  but  speakers  do  not  look  upon  them  as  differential 

Nasals  are  put  inside  parentheses  in  the  table;  although  the 



mouth  is  completely  closed  during  their  articulation,  the  opening 
of  the  nasal  channel  gives  them  wider  aperture  (see  Class  C). 

B.  Aperture  1:  Fricatives  or  Spirants 

The  phonemes  of  Class  B  are  characterized  by  incomplete  closure 
which  allows  the  air  to  pass  through  the  oral  cavity.  The  name 
spirant  is  all  too  general;  while  the  word  fricative  tells  nothing 
about  the  degree  of  closure,  it  does  suggest  friction  resulting  from 
the  expulsion  of  air  (Latin  fricdre) . 

The  phonemes  of  Class  B,  unlike  those  of  Class  A,  do  not  fall  into 
three  types.  First,  labials  proper  (corresponding  to  p  and  b)  are 
rarely  used ;  I  shall  disregard  them ;  they  are  ordinarily  replaced  by 
labiodentals,  which  are  produced  by  contact  between  the  lower  lip 
and  upper  teeth  (/  and  v).  Dentals  are  divided  into  several  va- 
rieties, depending  on  the  shape  which  the  tip  of  the  tongue  takes 
on  making  contact;  without  going  into  detail,  I  shall  use  jS,  0',  and 
j8"  to  designate  the  different  shapes  of  the  tip  of  the  tongue.  Among 
the  sounds  that  involve  the  palate,  the  ear  generally  singles  out  a 
front  articulation  (palatal)  and  a  back  articulation  (velar)  ^ 












[     ] 
[     ] 


[      ] 


[     ] 
[     ] 


[     ] 


[     ] 
[     ] 


[     ] 


[     ] 
[     ] 


[     ] 

J)  =  English  th  in  thing 
t5  =        "       th  in  then 
8  =        "        s  in  say 
s  in  rose 
sh  in  show 
i  =        "        g  in  rouge 
x'  =  German  ch  in  ich 
y'  =  North  German  g  in  liegen 
X  =  German  ch  in  Bach 
7  =  North  German  g  in  Tage 

*  Faithful  to  his  method,  Saussure  did  not  think  it  necessary  to  make  the 
same  distinction,  for  Class  A,  in  spite  of  the  importance  of  the  two  series 
Ki  and  K2  in  Proto-Indo-European.  The  omission  is  deliberate.  [Ed.] 








[      ] 


[    ] 


[      ] 
[      ] 


[      ] 



Is  there  a  sound  among  the  fricatives  to  match  n,  m,  n,  etc. 
among  the  occlusives — i.e.  a  nasal  v,  z,  etc.?  It  is  easy  to  imagine 
that  there  is;  for  instance,  a  nasal  v  is  heard  in  French  inventer 
'invent';  but  in  most  languages  the  nasal  fricative  is  not  a  dis- 
tinctive sound. ^ 

C.  Aperture  2:  Nasals  (see  above,  p.  46) 

D.  Aperture  3:  Liquids 

Two  kinds  of  articulation  are  classed  as  liquids. 

(1)  In  lateral  articulation  (indicated  by  I  in  the  formulas  below) 
the  tongue  rests  against  the  front  palate  but  leaves  an  opening  on 
both  sides.  It  is  possible  to  single  out,  according  to  the  place  of 
articulation,  dental  /,  palatal  V ,  and  guttural  of  velar  I.  In  most 
languages  lateral  phonemes  are  voiced  in  the  same  way  as  6,  z,  etc. 
Still,  a  voiceless  lateral  is  not  impossible ;  it  exists  even  in  French, 
where  an  I  that  follows  a  voiceless  phoneme  may  be  pronounced 
without  the  laryngeal  sound  (e.g.  the  I  of  pluie  'rain'  against  the 
I  of  bleu  'blue') ;  but  speakers  are  not  conscious  of  the  difference. 

There  is  no  point  in  discussing  nasal  /,  which  is  very  rare  and 
nondifferentiating,  although  it  does  occur,  especially  after  a  nasal 
sound  (e.g.  the  I  in  French  branlant  'shaking'). 

(2)  In  vibrant  articulation  (indicated  by  v  in  the  formula  below) 
the  tongue  is  held  farther  from  the  palate  than  for  I,  but  a  variable 
number  of  contacts  between  the  tongue  and  palate  makes  the 
aperture  for  vibrants  equivalent  to  the  aperture  for  laterals. 
Vibration  is  produced  in  two  ways:  with  the  tip  of  the  tongue 
thrust  forward  against  the  alveolar  ridge  (trilled  r),  or  with  the 
back  of  the  tongue  in  contact  with  the  palate  (a  dorsal  r  or  burr) . 
What  was  said  about  voiced  or  nasal  laterals  is  also  applicable  to 






[     ] 


[  ] 


[     ] 


[     ] 


[     ] 

*  The  French  reads,  ''mais  en  g6n(5rale  la  fricative  nasale  n'est  pas  un  son, 
dont  la  langue  ait  conscience."  [Tr.] 


Beyond  aperture  3,  we  enter  into  a  new  field ;  from  consonants  we 
pass  to  vowels.  Up  to  this  point,  I  have  not  brought  up  the  dis- 
tinction between  the  two  for  a  very  simple  reason :  the  phonational 
mechanism  is  the  same  for  both.  The  formula  for  a  vowel  is  com- 
parable in  every  way  to  the  formula  for  a  voiced  consonant.  From 
the  viewpoint  of  oral  articulation,  no  distinction  need  be  made. 
Only  the  acoustical  effect  is  different.  Beyond  a  certain  degree  of 
aperture,  the  mouth  functions  mainly  as  a  resonator:  the  timbre 
of  the  laryngeal  sound  stands  out,  and  oral  noise  decreases.  How 
much  of  the  larjmgeal  sound  is  cut  out  depends  on  how  tightly  the 
mouth  is  closed;  the  wider  the  mouth  is  opened,  the  more  noise 
lessens;  thus  sound  predominates  in  vowels  through  a  purely 
mechanical  process. 

E.  Aperture  4:  h  u,  ii 

The  vowels  of  Class  E  require  much  more  closure  than  the  other 
vowels — almost  as  much  as  consonants.  Certain  consequences  that 
will  appear  later  justify  the  name  semi- vowels,  which  is  generally 
given  to  phonemes  of  Class  E. 

The  phoneme  i  is  pronounced  with  retracted  lips  ( — )  and  front 
articulation,  u  with  rounded  hps  (O)  and  back  articulation,  and 
ii  with  the  lip  position  of  u  and  the  articulation  of  i. 

Like  all  other  vowels,  i,  u,  and  ii  have  nasaUzed  forms.  Here 
we  can  disregard  them  since  they  are  rare.  It  is  worth  noting, 
however,  that  the  sounds  written  in  and  un  in  French  are  really  not 
nasalized  i  and  u  (see  below). 

Is  there  a  voiceless  i,  i.e.  articulated  without  a  laryngeal  sound? 
The  same  question  arises  for  u  and  iX,  and  for  all  vowels.  Such 
phonemes,  corresponding  to  voiceless  consonants,  exist  but  are 
not  to  be  confused  with  whispered  vowels,  i.e.,  vowels  articulated 
with  the  glottis  relaxed.  Voiceless  vowels  are  like  the  aspirated  /I's 
that  are  pronounced  before  them :  in  hi,  an  i  with  no  vibration  is 
first  heard,  then  a  normal  i. 

F.  Aperture  5:  e,  o,  6 



The  articulation  of  the  phonemes  of  Class  F  corresponds  exactly 
to  the  articulation  of  i,  u,  ii.  Nasalized  vowels  occur  frequently 
(e.g.  French  e,  6,  o  as  in  pin  'pine,'  pont  'bridge/  brun  'brown'). 
Voiceless  forms  are  the  aspirated  h  of  he,  ho,  ho. 

N.  B.  Many  languages  single  out  several  degrees  of  aperture 
within  Class  F;  French,  for  instance,  has  at  least  two  series,  one 
closed  (e,  o,  o  as  in  de  'thimble,'  dos  'back,'  deux  'two')  and  the 
other  open  (e,  p,  p  as  in  mer  'sea,'  mort  'death,'  meurtre  'murder'). 







— 5f 

[     ] 


[     ] 


[     ] 

— 5f 



— 6i  — 6i 

t     J  •  •  •  •  the  h  of  ha 

G.  Aperture  6:  a 

The  a  has  maximum  aperture.  This  vowel 
has  a  nasaUzed  form,  a — slightly  more  con- 
tracted,  to  be  sure — and  a  voiceless  form, 

Chapter  II 

1.  Need  for  Studying  Sounds  in  the  Spoken  Chain 

Detailed  analyses  of  speech  sounds  can  be  found  in  special 
treatises,  especially  in  the  works  of  English  phoneticians. 

Do  detailed  analyses  alone  fulfill  the  auxiliary  role  of  phonology 
in  the  science  of  linguistics?  Such  a  mass  of  details  has  no  value  in 
itself;  only  synthesis  matters.  The  linguist  does  not  need  to  be  a 
consummate  phonologist ;  he  asks  only  to  be  given  certain  data 
that  are  necessary  for  the  study  of  language. 

The  method  of  phonology  is  particularly  faulty  at  one  point: 
phonologists  too  often  forget  that  language  is  made  up  not  only  of 


sounds  but  also  of  expanses  of  spoken  sounds;  they  still  do  not 
devote  enough  attention  to  the  reciprocal  relations  of  sounds. 
These  relations  are  not  immediately  discernible;  syllables  are 
easier  to  identify  than  their  sounds.  We  have  seen  (pp.  25  ff.)  that 
some  primitive  systems  of  writing  noted  syllabic  units;  only  later 
was  the  alphabetic  system  devised. 

Besides,  it  is  never  a  simple  unit  that  proves  embarrassing  in 
linguistics.  If  at  a  particular  moment  every  a  became  o  in  a  par- 
ticular language,  nothing  would  result  from  the  change;  the  lin- 
guist may  simply  record  the  phenomenon  without  trying  to  explain 
it  phonologically.  The  science  of  sounds  becomes  invaluable  only 
when  two  or  more  elements  are  involved  in  a  relationship  based 
upon  their  inner  dependence,  for  the  variations  of  each  element 
are  limited  by  the  variations  of  the  other  element  or  elements ;  the 
single  fact  that  there  are  two  elements  calls  for  a  relationship  and 
a  rule — and  this  is  quite  different  from  a  simple  statement.  In 
trying  to  find  a  phonological  principle,  this  science  is  then  contra- 
dicting itself  by  showing  partiality  to  isolated  sounds.  Two  pho- 
nemes are  enough  to  lead  to  bewilderment.  In  Old  High  German, 
for  instance,  hagl,  balg,  wagn,  lang,  donr,  dorn  later  became  hagal, 
halg,  wagan,  lang,  donnar,  dorn;  the  result  differs  according  to  the 
nature  and  the  order  of  the  phonemes  involved ;  sometimes  a  vowel 
occurs  between  the  original  consonants,  sometimes  the  combina- 
tion is  left  intact.  But  how  can  the  law  be  formulated?  Where  did 
the  difference  originate?  Doubtless  in  the  combinations  of  the  con- 
sonants {gl,  Ig,  gn,  etc.)  contained  in  the  words.  Each  combination 
obviously  contains  an  occlusive  that  is  either  preceded  or  followed 
by  a  liquid  or  a  nasal.  But  what  does  that  prove?  As  long  as  we  look 
upon  g  and  n  as  homogeneous  quantities,  we  cannot  understand 
why  the  mere  order  of  contact  in  g-n  and  n-g  should  affect  the 

Beside  the  phonology  of  species,  there  is  then  room  for  a  com- 
pletely different  science  that  uses  binary  combinations  and  se- 
quences of  phonemes  as  a  point  of  departure,  and  this  is  something 
else  entirely.  In  the  study  of  isolated  sounds,  to  note  the  position 
of  the  vocal  organs  is  sufficient ;  the  acoustical  quality  of  a  phoneme 
is  not  an  issue,  for  it  is  determined  by  the  ear;  as  for  articulation, 


the  speaker  has  unlimited  freedom.  But  when  we  come  to  the 
pronunciation  of  two  sounds  that  are  joined,  the  problem  is  not 
so  simple ;  we  must  bear  in  mind  the  possible  discrepancy  between 
the  effect  desired  and  the  effect  produced.  We  do  not  alwaj^s  have 
the  ability  to  pronounce  what  we  intend.  Freedom  in  linking  pho- 
nological species  is  checked  by  the  possibility  of  linking  articu- 
latory  movements.  To  give  an  account  of  what  takes  place  within 
groups,  there  should  be  a  science  of  sound  that  would  treat  articu- 
latory  movements  like  algebraic  equations:  a  binary  combination 
implies  a  certain  number  of  mechanical  and  acoustical  elements 
that  mutually  condition  each  other;  the  variation  of  one  has  a 
necessary  and  calculable  repercussion  on  the  others.  In  a  pho- 
national  act,  the  one  thing  which  has  a  universal  character  that 
places  it  above  all  the  local  differences  of  its  phonemes  is  the 
mechanical  regularity  of  the  articulatory  movements.  The  impor- 
tance of  combinatory  phonology  in  general  linguistics  is  obvious. 
Whereas  traditional  phonology  generally  gives  rules  for  articulat- 
ing all  sounds — variable  and  accidental  elements  of  languages — 
and  stops  there,  combinatory  phonology  limits  the  possibilities  and 
defines  the  constant  relations  of  interdependent  phonemes.  The 
case  of  hagl,  balg,  etc.  (see  p.  50)  brings  up  the  much  discussed 
question  of  Proto-Indo-European  sonants;  now  combinatory  pho- 
nology is  most  helpful  in  resolving  the  question,  for  the  syllabic 
grouping  of  phonemes  is  its  sole  concern  from  start  to  finish. 
Though  that  is  not  the  only  problem  to  be  solved  by  the  same 
method,  one  fact  is  certain;  we  simply  cannot  discuss  the  question 
of  sonants  unless  we  give  full  consideration  to  the  laws  that  govern 
the  combining  of  phonemes. 

2.  Implosion  and  Explosion 

I  shall  start  from  a  basic  observation:  there  is  a  perceptible 
difference  in  the  pronunciation  of  the  two  p's  of  appa.  The  first  p 
results  from  closure,  the  second  from  release.  The  two  impressions 
are  so  similar  that  phoneticians  used  a  single  p  to  transcribe  the 
sequence  pp  (see  p.  41,  note).  But  we  can  use  special  signs  (><)  to 
indicate  this  difference  between  the  two  p's  of  appa  (appa)  and  to 
identify  them  when  they  do  not  follow  each  other  (cf .  apta,  aipa) . 


This  distinction  holds  for  all  other  occlusives  and  for  fricatives 
{ajja) ,  nasals  (aThma) ,  liquids  (alia) ,  and  for  all  phonemes  in  general, 
including  all  vowels  except  a{a6oa). 

Closure  has  been  termed  implosion  and  release  explosion.  A  p  is 
either  implosive  (p)  or  explosive  (p).  We  may  speak  in  the  same 
sense  of  closing  and  opening  sounds. 

Doubtless  we  can  single  out,  besides  implosion  and  explosion,  an 
interval  during  which  occlusion  is  prolonged  at  will;  and  if  a 
phoneme  has  wider  aperture  (cf.  the  I  of  alia)  the  emission  of  the 
sound  itself  continues  while  the  vocal  organs  remain  motionless. 
Generally,  all  spoken  chains  contain  intermediate  stretches  that  I 
shall  call  holds  or  sistants.  But  they  are  like  implosive  articulations, 
for  their  effect  is  the  same.  In  the  following  pages  I  am  going  to 
consider  only  implosions  and  explosions.® 

The  method  I  have  outlined  would  be  unacceptable  in  a  com- 
prehensive treatment  of  phonology,  but  it  is  justifiable  in  a  sketch 
designed  to  reduce  the  essentials  of  syllabication  to  as  simple  a  plan 
as  possible.  I  do  not  pretend  to  resolve  thereby  all  the  difficulties 
brought  about  by  dividing  the  spoken  chain  into  syllables,  but 
simply  to  provide  a  rational  basis  for  studying  the  problem. 

One  further  remark.  Opening  and  closing  movements  necessary 
for  the  emission  of  sounds  must  not  be  confused  with  the  different 
apertures  of  the  sounds  themselves.  Any  given  phoneme  can  be 
both  implosive  and  explosive,  but  aperture  does  not  influence 
implosion  and  explosion  in  the  sense  that  the  two  movements  be- 
come less  distinct  as  aperture  increases.  In  i,  u,  u  the  difference  is 
still  quite  apparent ;  in  alia  we  can  detect  a  closing  i  and  an  opening 
i:  similarly,  in  auua,  auua  the  implosive  sound  and  the  following 
explosive  sound  differ  so  sharply  that  writing  sometimes  breaks  its 
regular  pattern  and  records  the  difference;  English  w,  German  j, 

^  Saussure's  treatment  of  holds  is  one  of  the  most  debatable  points  in  his 
theory.  To  prevent  certain  objections  one  should  note  that  any  sistant  (e.g. 
that  in  the  articulation  of/)  is  the  result  of  two  forces:  (1)  the  pressure  of  air 
against  the  opposing  organs  and  (2)  the  resistance  of  the  organs  as  they  tighten 
to  equalize  the  pressure.  A  hold  is  thus  only  continued  implosion.  That  is  why 
the  effect  is  the  same  throughout  whenever  a  hold  and  an  implosive  sound  of 
the  same  species  are  uttered  in  sequence.  Accordingly,  to  unite  the  two  types 
of  articulation  in  one  mechanical  and  acoustical  entity  is  not  illogical. 
Explosion,  on  the  contrary,  is  opposed  to  both:  by  definition  it  is  a  release. 
See  also  Section  6.  [Ed.i 


and  often  French  y  (in  yeux  'eyes/  etc.)  stand  for  opening  sounds 
in  opposition  to  u  and  i,  which  are  used  for  u  and  I.  But  when  the 
aperture  is  wider  (e.g.  e  and  o)  it  is  hardly  possible  to  distinguish 
between  implosion  and  explosion  in  practice,  although  a  difference 
is  theoretically  conceivable  (cf.  aeea,  aboa).  Finally,  as  we  have 
already  seen,  maximum  aperture  wipes  out  all  difference;  a  has 
neither  implosion  nor  explosion. 

The  table  of  phonemes  must  therefore  be  redoubled,  except  for 
a,  and  the  following  list  of  irreducible  units  set  up : 



















Far  from  discarding  the  distinctions  sanctioned  by  spelling  (il,  I), 
I  shall  carefully  preserve  them  (w,  y) ;  justification  for  my  view- 
point will  be  found  below  (see  Section  7), 

For  the  first  time  we  have  broken  away  from  abstraction.  Now 
for  the  first  time  we  have  found  the  concrete,  irreducible  units  that 
occupy  a  place  and  correspond  to  a  beat  in  the  spoken  chain :  p  was 
nothing  except  an  abstract  unit  linking  the  common  characters  of 
p  and  p,  the  only  units  that  actually  exist.  In  the  same  way,  the 
still  higher  abstraction  of  "labiahty"  links  together  P  B  M.  We 
may  speak  of  P  as  if  it  were  a  zoological  species;  there  are  male  and 
female  representatives  of  the  species,  but  there  is  no  ideal  specimen. 
Before,  we  had  been  singling  out  and  classifying  the  abstractions; 
but  we  had  to  go  beyond  the  abstract  to  reach  the  concrete. 

Phonology  made  a  great  mistake  in  considering  abstractions  real 
units  without  examining  more  carefully  the  definition  of  the  unit. 
The  Greek  alphabet  was  successful  in  singling  out  the  abstract 
elements — an  accomplishment  that  presupposes  a  most  remark- 
able analysis  (see  p.  39) ;  still,  the  analysis  of  the  Greeks  was  in- 
complete, for  it  was  not  carried  out  fully. 

Exactly  what  is  an  unquaUfied  p?  Considered  in  time  as  part  of 
the  spoken  chain,  it  is  neither  specifically  p  nor  p,  and  still  less  pp, 


this  combination  being  clearly  decomposable ;  and  if  we  consider  it 
outside  the  spoken  chain,  it  is  a  thing  which  has  no  independent 
existence  and  with  which  we  can  do  nothing.  By  itself,  what  does 
a  combination  like  I  +  g  mean?  Two  abstractions  cannot  form  a 
moment  in  time.  But  to  talk  about  Ik,  tic,  Ik,  tk,  and  thus  to  draw 
together  the  genuine  elements  of  speaking  is  quite  different.  Then 
we  see  why  two  elements  suffice  to  embarrass  traditional  pho- 
nology, and  the  impossibility  of  working  with  abstract  phonological 
units — as  it  did — is  demonstrated. 

One  theory  states  that  in  any  simple  phoneme  considered  in  the 
chain  (e.g.,  p  in  pa  or  aya),  implosion  and  explosion  {dpa)  occur 
successively.  Doubtless  any  release  must  be  preceded  by  closure. 
To  take  still  another  example,  in  pronouncing  fp  I  must  first  estab- 
lish closure  for  r,  then  articulate  an  opening  r  while  closure  for  p  is 
being  formed  by  the  lips.  But  I  need  only  specify  my  viewpoint  in 
order  to  answer  that  objection.  In  analyzing  a  phonational  act,  I 
shall  consider  only  the  differential  elements  that  make  a  distinct 
impression  on  the  ear,  allowing  delimitation  of  the  acoustical  units 
of  the  spoken  chain.  Only  the  acoustic-motor  units  are  to  be  con- 
sidered; hence  the  articulation  of  explosive  r  along  with  implosive 
p  is  nonexistent  to  me,  for  it  produces  no  perceptible  sound,  or  at 
least  is  not  important  in  the  chain  of  phonemes.  One  must  appreci- 
ate this  basic  point  fully  in  order  to  understand  the  developments 
that  follow. 

3.  Different  Combinations  of  Explosions  and  Implosions  in  the  Chain 
Consider  now  what  may  result  from  each  sequence  of  the  four 
combinations  of  implosives  and  explosives  that  are  theoretically 
possible:  (1)  <>,  (2)  ><,  (3)  «,  (4)  ». 

1)  Explosive-Implosive  Combination  (<>).  Without  breaking 
the  spoken  chain,  we  can  always  join  explosive  and  implosive  pho- 
nemes: kf,  pi,  yrh,  etc.  (e.g.  Sanskrit  kfta-,  Enghsh  pity,  Proto- 
Indo-European  *yrhto-,  etc.).  Of  course,  some  combinations  hke  kl, 
etc.  have  no  practical  acoustical  effect,  but  the  fact  remains  that 
the  articulating  of  an  opening  k  leaves  the  vocal  organs  in  the  right 
position  for  making  closure  at  any  given  point.  The  two  pho- 
national movements  do  not  interfere  with  each  other. 

2)  Implosive-Explosive  Combination  (><).     Under  the  same  con- 


ditions — and  with  the  same  reservations — it  is  always  possible  to 
join  implosive  and  explosive  phonemes:  im,  Jet,  etc.  (e.g.  Greek 
haima,  English  active,  etc.). 

Of  course  the  successive  articulatory  moments  do  not  follow  each 
other  so  naturally  as  they  do  in  the  reverse  order  of  combination  1. 
The  difference  between  initial  implosions  and  explosions  is  this: 
explosion,  which  tends  to  neutralize  the  vocal  organs,  does  not 
engage  the  following  moment;  but  implosion  sets  up  a  definite 
position  that  cannot  be  the  point  of  departure  for  just  any  ex- 
plosion. For  that  reason  one  must  always  resort  to  some  facilitating 
movement  to  put  the  organs  necessary  for  articulating  the  second 
phoneme  into  the  right  position.  While  executing  s  in  sp,  for 
instance,  the  hps  must  close  to  prepare  for  opening  p.  But  ex- 
perience shows  that  the  facilitating  movement  has  no  appreciable 
effect.  It  produces  only  a  furtive  sound  that  in  no  way  interferes 
with  the  succession  of  the  chain. 

3)  Implosive  Link  («) .  Two  consecutive  explosions  can  be  pro- 
duced, but  if  the  second  belongs  to  a  phoneme  of  less  or  of  equal 
aperture,  the  impression  of  acoustical  unity  that  results  in  the 
opposite  case  or  in  the  sequences  of  combinations  1  and  2  will  be 
missing:  pk  can  be  pronounced  (pka),  but  these  sounds  do  not  form 
a  chain,  for  the  P  and  K  species  have  the  same  aperture.  This 
rather  unnatural  pronunciation  would  result  from  stopping  after 
the  first  a  of  cha-pka.''  On  the  contrary,  pr  gives  the  impression  of 
continuity  (cf .  price) ;  nor  does  fy  cause  difficulty  (cf .  French  rien 
'nothing').  Why?  Because  at  the  very  instant  the  first  explosion 
occurs,  the  vocal  organs  have  already  assumed  the  right  position 
for  executing  the  second  explosion  without  interfering  with  the 
acoustical  effect  of  the  first;  thus  the  organs  are  already  in  position 
for  the  r  of  price  while  p  is  being  pronounced.  But  it  is  impossible 
to  pronounce  the  reverse  series  rp,  not  because  this  is  mechanically 
impossible  (we  can  prepare  for  p  while  articulating  opening  f),  but 
because  the  movement  of  the  f,  coming  against  the  smaller  aper- 
ture of  p,  would  be  imperceptible.  Two  separate  movements  would 

^  To  be  sure  combinations  of  explosive  phonemes  having  the  same  aperture 
are  very  common  in  some  languages  (e.g.  initial  kt  in  Greek;  cf.  ktelno);  al- 
though these  combinations  are  easy  to  pronounce,  they  lack  acoustical  unity. 
(See  the  following  note.) 


be  required  to  make  fp  audible,  and  the  emission  would  be  inter- 

A  continuous  explosive  link  may  include  more  than  two  elements 
provided  that  each  successive  aperture  is  wider  than  the  preceding 
one  (e.g.  kfwa).  Aside  from  a  few  special  cases  which  I  cannot  dis- 
cuss in  detail,^  the  natural  limit  of  the  possible  number  of  ex- 
plosions is  the  number  of  degrees  of  aperture  distinguishable  in 

4)  Implosive  Link  (>>).  The  reverse  law  governs  the  implosive 
link :  whenever  a  particular  phoneme  is  more  open  than  the  f ollow-_ 
ing  one,  the  impression  of  continuity  persists  (e.g.  if,  ft);  if  this 
condition  is  not  met — if  the  following  phoneme  is  more  open  or  has 
the  same  aperture — pronunciation  is  still  possible,  but  the  impres- 
sion of  continuity  is  lacking :  sf  in  dsfta  is  basically  the  same  as  pJc 
in  cha-pka  (see  p.  55).  This  phenomenon  parallels  the  one  an- 
alyzed in  the  explosive  link  in  every  way :  in  ft  the  i,  by  virtue  of  its 
narrower  aperture,  exempts  r  from  explosion;  in  a  link  like  fm, 
made  up  of  phonemes  with  different  points  of  articulation,  rh  does 
not  exempt  f  from  exploding  but  brings  about  the  same  result  by 
covering  its  explosion  completely.  Otherwise,  as  in  the  reverse 
order  mf,  the  furtive,  mechanically  indispensable  explosion  breaks 
the  spoken  chain. 

An  implosive  link,  hke  an  explosive  one,  obviously  can  include 
more  than  two  elements  if  each  has  wider  aperture  than  the  follow- 
ing one  (cf.  dfst). 

Leaving  aside  the  breaking  of  links,  we  turn  now  to  the  normal 
continuous  chain — one  that  might  be  termed  physiological — as  rep- 

*  Through  dehberate  over-simpUfication,  Saussure  considers  here  only  the 
degree  of  aperture  of  the  phoneme,  not  the  place  and  specific  nature  of  its 
articulation  (whether  voiceless  or  voiced,  vibrant  or  lateral,  etc.)-  Conclusions 
drawn  from  the  principle  of  aperture  alone  are  not  applicable  without  exception 
to  all  actual  cases.  In  a  sequence  like  trya,  for  instance,  only  with  difficulty  can 
the  first  three  elements  be  pronounced  without  breaking  the  chain:  ifyd 
(unless  y  palatalizes  the  r  and  merges  with  it) ;  but  the  three  elements  in  try 
make  a  perfect  explosive  link  (cf.  also  p.  63  concerning  meurtrier,  etc.);  trwa, 
on  the  contrary,  offers  no  difficulty.  Links  like  pmla,  etc.,  where  it  is  difficult 
to  avoid  pronouncing  the  nasal  implosively  (pmld),  should  also  be  cited.  The 
aberrant  cases  show  up  especially  in  explosion,  an  instantaneous  act  that 
tolerates  absolutely  no  hindrances.  [Ed.] 


resented  by  French  particuUerement:  pdfiikulyerrhd.^ 
The  chain  is  characterized  by  a  succession  of  graduated  links  cor- 
responding to  a  succession  of  releases  and  closures  of  the  vocal 

The  normal  chain  thus  defined  makes  possible  the  following 
observations  which  are  of  capital  importance. 

4.  Syllabic  Boundary  and  Vocalic  Peak 

Passing  from  an  implosion  to  an  explosion  in  a  chain  of  sounds 
_produces  a  peculiar  effect  that  marks  the  syllabic  boundary  (e.g. 
the  ik  of  particuUerement) .  The  regular  coincidence  of  a  mechanical 
principle  and  a  definite  acoustical  effect  assures  the  implosive- 
explosive  combination  of  a  right  to  existence  in  phonology.  Its 
character  persists  regardless  of  the  species  that  compose  it.  It 
constitutes  a  type  that  contains  as  many  species  as  there  are 
possible  combinations. 

The  syllabic  boundary  sometimes  occurs  at  different  points  in 
the  same  series  of  phonemes,  depending  on  the  speed  of  passage 
from  implosion  to  explosion.  In  ardra,  for  instance,  neither  the 
division  dfdfd  nor  the  division  dfdfd  breaks  the  chain,  for  both  the 
implosive  link  dfd  and  the  explosive  link  df  are  graduated.  The 
same  would  apply  to  Ulye  of  particuUerement  (ulye  or  ulye) . 

Next,  we  notice  that  in  passing  from  silence  to  initial  implosion 
(>) — e.g.  art  in  artist— or  from  explosion  to  implosion  (<>)  e.g.  part 
in  particuUerement — the  sound  where  the  initial  implosion  occurs 
is  distinguished  from  neighboring  sounds  by  its  own  vocalic  effect. 
In  no  way  does  the  vocalic  effect  depend  on  the  wider  aperture  of 
the  sound  a,  for  in  pft,  r  produces  the  same  effect;  it  is  inherent  in 
initial  implosions  regardless  of  their  phonological  species,  i.e.,  their 
degree  of  aperture ;  whether  the  implosion  comes  after  a  silence  or 
after  an  explosion  matters  little.  A  sound  that  makes  a  vocalic 
impression  is  a  vocalic  peak. 

Vocalic  peaks  have  also  been  called  sonants,  and  all  other  sounds 
in  the  same  syllable  con-sonants  [consonantcs].  Vowels  and  con- 
sonants [consonnes]  designate  different  species  (see  p.  48) ;  sonants 

^  Note  the  difference  in  the  syllabication  of  English  particularly  [par  tik 
iu  iaf  ii].  [Tr.] 


and  con-sonants,  on  the  other  hand,  designate  functions  within 
syllables.  The  dual  system  of  terminology  clears  up  the  confusion 
that  has  existed  for  a  long  time.  Thus  the  /  species  is  the  same  in 
French  fidele  'faithful'  and  pied  'foot';  it  is  a  vowel ;^°  but  it  is  a 
sonant  in  fidele  and  a  con-sonant  in  pied.  Analysis  shows  that 
sonants  are  always  implosive  while  non-sonants  may  be  either 
implosive  (e.g.  I  in  English  hoi,  written  hoy)  or  explosive  (e.g.  y  in 
French  fye,  written  pied).  Analysis  only  confirms  the  distinction 
set  up  between  the  two  classes.  Regularly,  e,  o,  a  are  sonants,  but 
this  is  merely  a  coincidence :  having  wider  aperture  than  any  of  the 
other  sounds,  they  are  always  at  the  beginning  of  an  implosive 
chain.  Conversely  occlusives,  which  have  minimum  aperture,  are 
always  con-sonants.  In  practice  phonemes  of  apertures  2,  3,  and  4 
(nasals,  liquids,  and  semivowels)  play  either  role,  depending  on 
contiguous  sounds  and  the  nature  of  their  articulation. 

5.  Criticism  of  Theories  of  Syllahication 

The  ear  perceives  syllabic  division  in  every  spoken  chain ;  it  also 
perceives  a  sonant  in  every  syllable.  One  can  accept  both  facts  and 
still  wonder  why  they  should  hold  true.  Different  explanations 
have  been  offered. 

1)  Noticing  that  some  phonemes  are  more  sonorous  than  others, 
some  scholars  have  tried  to  base  syllables  on  the  sonority  of  pho- 
nemes. But  how  is  it  that  sonorous  phonemes  like  i  and  u  do  not 
necessarily  form  syllables?  Besides,  where  does  sonority  stop  since 
fricatives  like  s  are  syllabic  (e.g.  pst)l  If  only  the  relative  sonority 
of  sounds  in  contact  is  at  stake,  how  can  one  explain  such  com- 
binations as  wl  (e.g.  Proto-Indo-European  *wlkos  'wolf'),  where 
the  least  sonorous  element  is  syllabic? 

2)  E.  Sievers  was  the  first  to  show  that  a  sound  classed  as  a 
vowel  does  not  necessarily  make  a  vocalic  impression  (e.g.  we  saw 
above,  p.  52  f.,  that  y  and  w  are  nothing  except  i  and  u) ;  but  one 
who  asks  why  a  sound  should  have  a  dual  function — or  a  dual  acous- 
tical effect,  for  "function"  means  just  that — is  given  this  reply : 
the  function  of  a  given  sound  depends  on  whether  the  sound  re- 
ceives the  "syllabic  accent." 

This  is  a  vicious  circle.  If  I  am  free  under  all  circumstances  to 
"Cf.  English /ee  [fij]  and  few  [fju].  [Tr.] 


place  the  syllabic  accent  that  creates  sonants  wherever  I  choose, 
then  the  accent  might  as  well  be  called  sonantic.  But  if  syllabic 
means  anything,  its  meaning  must  derive  from  the  laws  of  the 
syllable.  Not  only  are  such  laws  lacking,  but  the  sonantic  quahty  is 
described  as  silbenbildend,  as  if  the  formation  of  syllables  depended 
on  syllabic  accent. 

The  difference  between  our  method  and  (1)  and  (2)  above  is 
obvious:  by  analyzing  syllables  as  they  occur  in  the  chain,  we  found 
the  irreducible  units,  opening  and  closing  sounds;  then  by  com-_ 
bining  these  units,  we  were  able  to  define  the  syllabic  boundary_ 
and  vocalic  peak.  Now  we  know  the  physiological  conditions  under 
which  the  acoustical  effects  must  occur.  The  theories  criticized 
above  follow  the  opposite  course:  from  isolated  phonological 
species,  the  proponents  of  the  theories  pretend  to  deduce  the 
boundary  of  the  syllable  and  the  position  of  the  sonant.  In  a  given 
series  of  phonemes,  one  pronunciation  may  be  more  natural  and 
easier  than  another;  but  by  and  large  the  possibility  of  choosing 
between  opening  and  closing  articulations  persists,  and  syl- 
labication depends  on  the  choice  rather  than  directly  on  phono- 
logical species. 

Doubtless  my  theory  neither  exhausts  nor  resolves  all  questions. 
Hiatus,  for  example,  which  occurs  very  frequently,  is  simply  a 
broken  implosive  link,  deliberate  or  unintentional:  e.g.  i-d  (in 
French  il  cria  'he  shouted')  and  d-l  (in  French  ebahi  'amazed')."  It 
occurs  more  easily  when  the  phonological  species  have  wide 

There  are  also  broken  explosive  links  which,  though  ungradu- 
ated,  fall  into  a  phonetic  chain  just  as  do  normal  groups.  I  men- 
tioned one  example  earlier,  kteino  (see  p.  55,  note).  Or  take  the 
sequence  -pzta:  normally  it  can  be  pronounced  only  pzta;  it  should 
comprise  two  syllables,  and  it  does  have  two  if  the  laryngeal  sound 
of  z  is  pronounced  distinctly;  but  if  z  is  muffled,  the  opposition 
between  it  and  a  is  insufficient  since  z  is  one  of  the  phonemes  that 
require  least  aperture;  the  result  is  that  only  one  syllable  is  per- 
ceived and  something  like  'pzld  is  heard. 

In  all  broken  explosive  finks,  when  wifi  and  intention  interfere, 
to  some  extent  it  wifi  be  possible  to  eschew  physiological  neces- 

"  Cf.  English  rearm  (z-d)  and  Aida  (d-t).  [Tr.] 


sities.  Determining  what  is  wilful  and  what  is  physiological  is  often 
difficult.  But  phonation  depends  on  a  succession  of  implosions  and 
explosions,  and  this  is  basic  in  syllabication. 

6.  Length  of  Implosion  and  Explosion 

Our  explanation  of  syllables  in  terms  of  the  functioning  of 
explosions  and  implosions  leads  to  an  important  observation  that 
is  simply  a  generalization  of  a  metrical  fact.  We  can  separate  two 
types  of  long  vowels  in  Latin  and  Greek:  those  long  by  nature 
(mater)  and  those  long  by  position  (f actus).  Why  is  fac  counted  long 
in  f actus?  because  of  the  at  combination?  No,  for  if  the  combination 
alone  determined  length,  every  syllable  beginning  with  two  con- 
sonants would  also  be  long;  but  this  is  not  true  (cf.  cliens,  etc.). 

The  real  reason  is  that  explosion  and  implosion  are  basically 
different  with  respect  to  length.  The  first  is  always  so  rapid  that 
it  cannot  be  measured  by  the  ear;  for  that  reason  also,  it  never 
makes  a  vocalic  impression.  Only  implosion  is  measurable;  hence 
we  feel  that  we  dwell  longer  on  the  vowel  where  implosion  begins. 

Besides,  we  know  that  vowels  which  occur  before  a  combination 
of  an  occlusive  or  fricative  and  a  liquid  are  treated  in  two  ways: 
the  a  in  patron  may  be  either  long  or  short ;  the  principle  is  the  same 
in  either  instance.  Actually  if  and  If  are  pronounced  with  equal 
ease ;  the  first  method  of  articulation  allows  a  to  remain  short ;  the 
second  creates  a  long  syllable.  The  same  dual  treatment  of  a  is  not 
possible  in  a  word  hke  f actus;  ci  can  be  pronounced,  but  ct  cannot. 

7.  Phonemes  of  Aperture  4;  Diphthongs;  Questions  about 

Finally,  the  phonemes  of  aperture  4  call  for  some  additional 
remarks.  We  have  seen  that,  contrary  to  what  happens  with  other 
sounds,  usage  has  sanctioned  a  double  set  of  graphs  (w  =  u,u  =  H; 
y  =  I,  i  =  i)  for  the  phonemes  of  aperture  4  (see  p.  53).  The  reason 
is  simple :  in  groups  like  aiya,  auwa  the  distinction  between  release 
and  closure  is  more  striking  than  elsewhere;  z  and  H  make  a  clear 
vocalic  impression,  I  and  il  a  consonantal  impression.^  Without 

^^  The  i  of  aperture  4  must  not  be  confused  with  the  soft  palatal  fricative 
(e.g.  the  g  in  North  German  liegen),  a  phonological  species  that  has  all  the 
characteristics  of  a  consonant.  [S.] 


pretending  to  explain  the  fact,  I  wish  to  point  out  that  consonantal 
i  is  never  accompanied  by  closure :  the  I  in  ai  never  has  the  same 
effect  as  the  y  in  aiya  (cf .  English  hoy  and  French  'pied) ;  through 
position,  then,  ?/  is  a  consonant  and  i  a  vowel,  for  these  variations 
of  the  I  species  do  not  occur  indifferently.  The  same  remarks  apply 
to  u  and  w,  ii  and  iv. 

The  preceding  discussion  clarifies  the  question  of  the  diphthong. 
It  is  only  a  special  kind  of  implosive  link;  dfta  and  duta  are  abso- 
lutely parallel;  only  the  aperture  of  the  second  element  is  different. 
A  diphthong  is  an  implosive  link  in  which  the  second  phoneme  is 
relatively  open,  making  a  specific  acoustical  impression.  We  might 
say  that  the  sonant  continues  in  the  second  element  of  the  com- 
bination. Conversely,  a  combination  like  iya  is  distinguished  from 
a  combination  like  tfa  only  by  the  degree  of  aperture  of  the  last 
explosive.  This  means  that  what  phonologists  call  ascending  diph- 
thongs are  not  really  diphthongs  but  explosive-im plosive  combina- 
tions in  which  the  first  element  does  not  produce  a  specific  acous- 
tical effect  even  though  it  is  relatively  open  (tyd).  Combinations 
like  uo,  !a,  with  the  accent  on  H  and  ^  (e.g.  buoh,  liab  in  certain 
German  dialects),  are  also  false  diphthongs  that  fail  to  make  the 
impression  of  unity  produced  by  6u,  di,  etc.;  we  cannot  pronounce 
Ho  as  implosive  +  implosive  and  avoid  breaking  the  link  with- 
out calhng  in  some  device  to  impose  an  artificial  unity  on  the 

Our  definition  of  the  diphthong — which  relates  it  to  the  general 
principle  of  implosive  links — shows  that  it  is  not,  as  one  might 
think,  an  incongruous  something  not  to  be  classed  among  phono- 
logical phenomena;  there  is  no  need  for  putting  it  into  a  special 
category.  The  uniqueness  of  the  diphthong  is  really  of  no  interest 
or  importance ;  the  important  thing  is  to  determine,  not  the  end  of 
the  sonant,  but  its  beginning. 

E,  Sievers  and  many  other  linguists  make  a  distinction  in  writing 
between  i,  u,  ii,  r,  rj,,  etc.  and  f,  u,  ii,  r,  n,  etc.  (}  =  unsilhisches  i, 
i  =  silhisches  i) ;  they  write  mirta,  niairta,  miarta  while  I  write 
mirta,  mairta,  myarta.  Having  noticed  that  i  and  y  belong  to  the 
same  phonological  species,  they  wanted  especially  to  have  a  single 
generic  sign  for  both  (still  clinging  to  the  notion  that  a  chain  of 
sound  is  composed  of  species  in  juxtaposition).  Their  transcription, 


though  based  on  oral  evidence,  is  illogical  and  eliminates  the  very 
distinction  that  should  be  made:  (1)  opening  i,u{=  y,  w)  are  con- 
fused with  closing  i,  u  (e.g.  they  cannot  distinguish  between  newo 
and  neuo) ;  conversely,  closing  i,  u  are  sliced  in  two  (cf .  mirta  and 
mairta).  Here  are  some  examples  of  difficulties  that  result  from 
using  Siever's  system.  First,  Old  Greek  dwis  and  duls  against  rhewd 
and  rheuma.  The  two  oppositions  occur  under  exactly  the  same 
phonological  conditions  and  are  usually  indicated  by  the  same 
graphic  symbols.  The  u  is  either  opening  (w)  or  closing  (u)  depend- 
ing upon  whether  the  following  phoneme  is  more  open  or  more 
closed.  But  the  transcription  duis,  duis,  rheuo,  rhey,ina  wipes  out 
completely  these  oppositions.  Similarly,  in  Proto-Indo-European 
the  two  series  mater,  mdtrai,  mater es,  mdtrsu  and  suneu,  sunewai, 
sunewes,  sunusu  are  strictly  parallel  in  their  dual  treatment  of  both 
r  and  u.  In  the  second  series  at  least,  the  opposition  between  implo- 
sives  and  explosives  is  crystal  clear  in  writing.  But  the  transcription 
that  I  have  criticized  {suneu,  suneuai,  suneues,  sunusu)  obscures 
the  opposition.  Existing  distinctions  between  opening  and  closing 
sounds  (u,  w,  etc.)  should  not  only  be  preserved  but  extended  to 
cover  the  whole  system.  Thus  we  should  write  mater,  mdtpai, 
mdtepes,  mdtrsu;  then  the  functioning  of  syllabication  would  stand 
out;  vocalic  peaks  and  syllabic  boundaries  would  be  revealed. 

Editor's  Note.  The  theories  discussed  above  throw  light  on 
several  problems,  some  of  which  Saussure  touched  upon  in  his 
lectures.  We  shall  give  a  few  examples. 

1)  Sievers  cites  heritrynnn  (German  herittenen)  as  a  typical 
example  to  show  that  a  single  sound  may  alternately  function  twice 
as  a  sonant  and  twice  as  a  non-sonant  (actually  n  functions  only 
once  as  a  con-sonant,  and  the  word  should  be  transcribed  heritrinn,, 
but  that  matters  little) .  No  example  would  show  more  clearly  that 
"sound"  and  "species"  are  not  synonymous.  For  if  we  dwell  on  the 
n,  i.e.  implosion  and  sistant  articulation,  the  result  is  only  a  long 
syllable.  To  create  an  alternation  of  sonantic  and  con-sonantic  n's, 
we  would  have  to  pass  from  implosion  (first  n)  to  explosion  (second 
n)  and  back  to  implosion  (third  n).  Since  the  two  implosions  are 
preceded  by  no  other  implosion,  both  are  sonantic. 

2)  In  French  words  like  meurtrier  'murderer,'  ouvrier  'worker/ 


etc.,  final  -trier,  -vrier  formed  only  one  syllable  regardless  of  how 
they  were  actually  pronounced  (cf.  p.  56,  note).  Later,  speakers 
began  to  pronounce  them  in  two  syllables  (meur-tri-er,  with  or 
without  hiatus,  i.e.  -IfU  or  ifiye).  The  change  was  brought  about, 
not  by  placing  a  "syllabic  accent"  on  the  i  element,  but  by  chang- 
ing its  explosive  articulation  to  implosive. 

The  \iilgar  pronunciation  of  ouvrier  is  ouverier}^  This  change  is 
similar  to  the  dividing  of  -vrier  into  two  syllables,  but  here  the 
second  element  (r)  rather  than  the  third  changed  its  articulation 
and  became  a  sonant :  uvfye  -^  uvfye.  An  e  subsequently  developed 
in  front  of  sonantic  r. 

3)  We  might  also  cite  the  well-known  case  of  prosthetic  vowels 
in  front  of  s  followed  by  a  consonant  in  French :  Latin  scutum  -^ 
iscutum  —^  French  escu,  ecu  'shield.'  Here  sk  is  a  broken  link 
(see  p.  55) ;  sk  is  more  natural.  But  implosive  s  serves  as  a  vocalic 
peak  when  at  the  beginning  of  the  sentence  or  when  the  preceding 
word  ends  in  a  consonant  with  weak  aperture.  Prosthetic  i  and  e 
only  exaggerate  the  sonantic  quality  of  s:  any  perceptible  phono- 
logical characteristic  tends  to  become  more  pronounced  whenever 
speakers  try  to  preserve  it.  The  same  phenomenon  is  responsible 
for  esclandre  'scandal'  and  the  vulgar  pronunciations  esquelette, 
estatue  (Standard  French  squelette  'skeleton,'  statue  'statue') ;  it  also 
shows  up  in  the  vulgar  pronunciation  of  the  preposition  de  'of,' 
transcribed  ed:  un  oeil  ed  tanche  'a  tench's  eye.'  Through  syncope 
de  tanche  became  d'tanche;  but  to  be  perceptible  in  this  position 
d  must  be  implosive  (dianche) ;  the  result  was  again  the  develop- 
ment of  a  prosthetic  vowel. 

4)  It  is  scarcely  necessary  to  come  back  to  Indo-European  so- 
nants and  to  ask,  for  example,  why  Old  High  German  hagl  changed 
to  hagal  while  balg  remained  intact.  Here  the  /,  the  second  element 
of  an  implosive  link  (bdlg),  functioned  as  a  con-sonant  and  had  no 
reason  to  change  its  function.  But  the  I  of  hagl,  also  implosive,  was 
a  vocalic  peak.  Being  sonantic,  it  developed  a  more  open  prosthetic 
vowel  (an  a  if  we  accept  spelling  as  evidence).  The  vowel  became 
less  distinct  with  the  passage  of  time,  however,  and  today  Hagel  is 

"  Cf.  English  burglar.  [Tr.] 


again  pronounced  hdgl  The  quality  of  the  I  is  responsible  for  the 
difference  between  the  pronunciation  of  the  German  word  and 
French  aigle  'eagle' :  Hagel  has  a  closing  /  while  the  French  word 
has  an  opening  /  followed  by  a  mute  e  (egh). 

General  Principles 

Chapter  I 

1.  Sign,  Signified,  Signifier 

Some  people  regard  language,  when  reduced  to  its  elements,  as 
a  naming-process  only — a  list  of  words,  each  corresponding  to  the 
thing  that  it  names.  For  example : 

This  conception  is  open  to  criticism  at  several  points.  It  assumes 
that  ready-made  ideas  exist  before  words  (on  this  point,  see  below, 
p.  Ill) ;  it  does  not  tell  us  whether  a  name  is  vocal  or  psychological 
in  nature  (arbor,  for  instance,  can  be  considered  from  either  view- 
point) ;  finally,  it  lets  us  assume  that  the  linking  of  a  name  and  a 
thing  is  a  very  simple  operation — an  assumption  that  is  anything 
but  true.  But  this  rather  naive  approach  can  bring  us  near  the 
truth  by  showing  us  that  the  linguistic  unit  is  a  double  entity,  one 
formed  by  the  associating  of  two  terms. 




We  have  seen  in  considering  the  speaking-circuit  (p.  11)  that 
both  terms  involved  in  the  linguistic  sign  are  psychological  and  are 



united  in  the  brain  by  an  associative  bond.  This  point  must  be 

The  linguistic  sign  unites,  not  a  thing  and  a  name,  but  a  concept 
and  a  sound-image.^  The  latter  is  not  the  material  sound,  a  purely- 
physical  thing,  but  the  psychological  imprint  of  the  sound,  the 
impression  that  it  makes  on  our  senses.  The  sound-image  is  sensory, 
and  if  I  happen  to  call  it  "material,"  it  is  only  in  that  sense,  and  by 
way  of  opposing  it  to  the  other  term  of  the  association,  the  concept, 
which  is  generally  more  abstract. 

The  psychological  character  of  our  sound-images  becomes  ap- 
parent when  we  observe  our  own  speech.  Without  moving  our  lips 
or  tongue,  we  can  talk  to  ourselves  or  recite  mentally  a  selection  of 
verse.  Because  we  regard  the  words  of  our  language  as  sound- 
images,  we  must  avoid  speaking  of  the  "phonemes"  that  make  up 
the  words.  This  term,  which  suggests  vocal  activity,  is  applicable 
to  the  spoken  word  only,  to  the  realization  of  the  inner  image  in 
discourse.  We  can  avoid  that  misunderstanding  by  speaking  of  the 
sounds  and  syllables  of  a  word  provided  we  remember  that  the 
names  refer  to  the  sound-image. 

The  linguistic  sign  is  then  a  two-sided  psychological  entity  that 
can  be  represented  by  the  drawing : 

The  two  elements  are  intimately  united,  and  each  recalls  the 
other.  Whether  we  try  to  find  the  meaning  of  the  Latin  word  arbor 
or  the  word  that  Latin  uses  to  designate  the  concept  "tree,"  it  is 

*  The  term  sound-image  may  seem  to  be  too  restricted  inasmuch  as  beside 
the  representation  of  the  sounds  of  a  word  there  is  also  that  of  its  articulation, 
the  muscular  image  of  the  phonational  act.  But  for  F.  de  Saussure  language  is 
essentially  a  depository,  a  thing  received  from  without  (see  p.  13).  The  sound- 
image  is  par  excellence  the  natural  representation  of  the  word  as  a  fact  of 
potential  language,  outside  any  actual  use  of  it  in  speaking.  The  motor  side  is 
thus  implied  or,  in  any  event,  occupies  only  a  subordinate  role  with  respect 
to  the  sound-image.  [Ed.] 



clear  that  only  the  associations  sanctioned  by  that  language  appeal 
to  us  to  conform  to  reality,  and  we  disregard  whatever  others 
might  be  imagined. 

Our  definition  of  the  linguistic  sign  poses  an  important  question 
of  terminology.  I  call  the  combination  of  a  concept  and  a  sound- 
image  a  sign,  but  in  current  usage  the  term  generally  designates 
only  a  sound-image,  a  word,  for  example  {arbor,  etc.).  One  tends 
to  forget  that  arbor  is  called  a  sign  only  because  it  carries  the  con- 
cept "tree,"  with  the  result  that  the  idea  of  the  sensory  part 
imphes  the  idea  of  the  whole. 

f   t 

Ambiguity  would  disappear  if  the  three  notions  involved  here 
were  designated  by  three  names,  each  suggesting  and  opposing  the 
others.  I  propose  to  retain  the  word  sign  [signe]  to  designate  the 
/"whole  and  to  replace  concept  and  sound-image  respectively  by 
\signi££d^  [signifie]  and  signifier  [signifiant] ;  the  last  two  terms  have 
the  advantage  of  indicatmg  the  opposition  that  separates  them 
from  each  other  and  from  the  whole  of  which  they  are  parts.  As 
regards  sign,  if  I  am  satisfied  with  it,  this  is  simply  because  I  do  not 
know  of  any  word  to  replace  it,  the  ordinary  language  suggesting 
no  other. 

The  linguistic  sign,  as  defined,  has  two  primordial  character- 
istics. In  enunciating  them  I  am  also  positing  the  basic  principles  of 
any  study  of  this  type. 


2.  Principle  I:  The  Arbitrary  Nature  of  the  Sign 

The  bond  between  the  signifier  and  the  signified  is  arbitrary. 
Since  I  mean  by  sign  the  whole  that  results  from  the  associating  of 
the  signifier  with  the  signified,  I  can  simply  say:  the  linguistic  sign 
is  arbitrary. 

The  idea  of  "sister"  is  not  linked  by  any  inner  relationship  to 
the  succession  of  sounds  s-6-r  which  serves  as  its  signifier  in  French ; 


that  it  could  be  represented  equally  by  just  any  other  sequence  is 
proved  by  differences  among  languages  and  by  the  very  existence 
of  different  languages:  the  signified  "ox"  has  as  its  signifier  6-6-/ 
on  one  side  of  the  border  and  o-k-s  (Ochs)  on  the  other. 

No  one  disputes  the  principle  of  the  arbitrary  nature  of  the  sign, 
but  it  is  often  easier  to  discover  a  truth  than  to  assign  to  it  its 
proper  place.  Principle  I  dominates  all  the  linguistics  of  language; 
its  consequences  are  numberless.  It  is  true  that  not  all  of  them  are 
equally  obvious  at  first  glance;  only  after  many  detours  does  one 
discover  them,  and  with  them  the  primordial  importance  of  the 

One  remark  in  passing:  when  semiology  becomes  organized  as 
a  science,  the  question  will  arise  whether  or  not  it  properly  includes 
modes  of  expression  based  on  completely  natural  signs,  such  as 
pantomime.  Supposing  that  the  new  science  welcomes  them,  its 
main  concern  will  still  be  the  whole  group  of  systems  grounded  on 
the  arbitrariness  of  the  sign.  In  fact,  every  means  of  expression  used 
in  society  is  based,  in  principle,  on  collective  behavior  or — what 
amounts  to  the  same  thing — on  convention.  Polite  formulas,  for 
instance,  though  often  imbued  with  a  certain  natural  expressive- 
ness (as  in  the  case  of  a  Chinese  who  greets  his  emperor  by  bowing 
down  to  the  ground  nine  times),  are  nonetheless  fixed  by  rule;  it  is 
this  rule  and  not  the  intrinsic  value  of  the  gestures  that  obliges  one 
to  use  them.  Signs  that  are  wholly  arbitrary  realize  better  thg,n  the 
others  the  ideal  of  the  semiological  process;  that  is  why  langu^^, 
the  most  complex  and^universal  of  all  systems  of  expression,  is  als(r 
the  most  characteristic;  in  this  sense  linguistics  can  become  the 
master-pattern  for  all  branches  of  semiology  although  language  is 
only  one  particular  semiological  system. 
f^~.!The  word  sywhol  has  been  used  to  designate  the  linguistic  sign, 
or  more  specifically,  what  is  here  called  the  signifier.  Principle  I  in 
particular  weighs  against  the  use  of  this  term.  One  characteristic 
of  the  symbol  is  that  it  is  never  wholly  arbitrary;  it  is  not  empty, 
for  there  is  the  rudiment  of  a  natural  bond  between  the  signifier 
and  the  signified.  The  symbol  of  justice,  a  pair  of  scales,  could  not 
be  replaced  by  just  any  other  symbol,  such  as  a  chariot. 

The  word  arbitrary  also  calls  for  comment.  The  term  should  not 


imply  that  the  choice  of  the  signifier  is  left  entirely  to  the  speaker 
(we  shall  see  below  that  the  individual  does  not  have  the  power  to 
change  a  sign  in  any  way  once  it  has  become  established  in  the 
linguistic  community) ;  I  mean  that  it  is  unmotivated,  i.e.  arbitrary 
in  that  it  actually  has  no  natural  connection  with  the  signified. 

In  concluding  let  us  consider  two  objections  that  might  be  raised 
to  the  establishment  of  Principle  I : 

1)  Onomatopoeia  might  be  used  to  prove  that  the  choice  of  the 
signifier  is  not  always  arbitrary.  But  onomatopoeic  formations  are 
never  organic  elements  of  a  linguistic  system.  Besides,  their  number 
is  much  smaller  than  is  generally  supposed.  Words  like  French 
fouet  'whip'  or  glas  'knell'  may  strike  certain  ears  with  suggestive 
sonority,  but  to  see  that  they  have  not  always  had  this  property 
we  need  only  examine  their  Latin  forms  (fouet  is  derived  from  fdgus 
'beech-tree,'  glas  from  dassimim  'sound  of  a  trumpet').  The  quahty 
of  their  present  sounds,  or  rather  the  quality  that  is  attributed  to 
them,  is  a  fortuitous  result  of  phonetic  evolution. 

As  for  authentic  onomatopoeic  words  (e.g.  glug-glug,  tick-tock, 
etc.),  not  only  are  they  limited  in  number,  but  also  they  are  chosen 
somewhat  arbitrarily,  for  they  are  only  approximate  and  more  or 
less  conventional  imitations  of  certain  sounds  (cf .  English  bow-bow 
and  French  ouaoua).  In  addition,  once  these  words  have  been  intro- 
duced into  the  language,  they  are  to  a  certain  extent  subjected  to 
the  same  evolution — phonetic,  morphological,  etc. — that  other 
words  undergo  (cf.  pigeon,  ultimately  from  Vulgar  Latin  plpio, 
derived  in  turn  from  an  onomatopoeic  formation) :  obvious  proof 
that  they  lose  something  of  their  original  character  in  order  to 
assume  that  of  the  linguistic  sign  in  general,  which  is  unmotivated. 

2)  Interjections,  closely  related  to  onomatopoeia,  can  be  at- 
tacked on  the  same  grounds  and  come  no  closer  to  refuting  our 
thesis.  One  is  tempted  to  see  in  them  spontaneous  expressions  of 
reality  dictated,  so  to  speak,  by  natural  forces.  But  for  most  inter- 
jections we  can  show  that  there  is  no  fixed  bond  between  their  sig- 
nified and  their  signifier.  We  need  only  compare  two  languages  on 
this  point  to  see  how  much  such  expressions  differ  from  one  lan- 
guage to  the  next  (e.g.  the  English  equivalent  of  French  ate!  is 
ouch!).  We  know,  moreover,  that  many  interjections  were  once 


words  with  specific  meanings  (of.  French  diable!  'darn!'  mordieu! 
'golly!'  from  mort  Dieu  'God's  death,'  etc.)-^ 

Onomatopoeic  formations  and  interjections  are  of  secondary 
importance,  and  their  symbolic  origin  is  in  part  open  to  dispute. 

3.  Principle  II:  The  Linear  Nature  of  the  Signifier 

The  signifier,  being  auditory,  is  unfolded  solely  in  time  from 
which  it  gets  the  following  characteristics :  (a)  it  represents  a  span, 
and  (b)  the  span  is  measurable  in  a  single  dimension;  it  is  a  line. 
While  Principle  II  is  obvious,  apparently  hnguists  have  always 
neglected  to  state  it,  doubtless  because  they  found  it  too  simple; 
nevertheless,  it  is  fundamental,  and  its  consequences  are  incal- 
culable. Its  importance  equals  that  of  Principle  I;  the  whole 
mechanism  of  language  depends  upon  it  (see  p.  122  f.).  In  contrast 
to  visual  signifiers  (nautical  signals,  etc.)  which  can  offer  simul- 
taneous groupings  in  several  dimensions,  auditory  signifiers  have 
at  their  command  only  the  dimension  of  time.  Their  elements  are 
presented  in  succession;  they  form  a  chain.  This  feature  becomes 
readily  apparent  when  they  are  represented  in  writing  and  the 
spatial  line  of  graphic  marks  is  substituted  for  succession  in  time. 
Sometimes  the  linear  nature  of  the  signifier  is  not  obvious.  When 
I  accent  a  syllable,  for  instance,  it  seems  that  I  am  concentrating 
more  than  one  significant  element  on  the  same  point.  But  this  is  an 
illusion ;  the  S3'^llable  and  its  accent  constitute  only  one  phonational 
act.  There  is  no  duality  within  the  act  but  only  different  op- 
positions to  what  precedes  and  what  follows  (on  this  subject,  see 
p.  131). 

*Cf.  English  goodness!  and  zounds!  (from  God's  wounds).  [Tr.] 


Chapter  II 

1.  Immutability 

The  signifier,  though  to  all  appearances  freely  chosen  with  re- 
spect to  the  idea  that  it  represents,  is  fixed,  not  free,  with  respect 
to  the  linguistic  community  that  uses  it.  The  masses  have  no  voice 
in  the  matter,  and  the  signifier  chosen  by  language  could  be  re- 
placed by  no  other.  This  fact,  which  seems  to  embody  a  contradic- 
tion, might  be  called  colloquially  "the  stacked  deck."  We  say  to 
language:  "Choose!"  but  we  add:  "It  must  be  this  sign  and  no 
other."  No  individual,  even  if  he  willed  it,  could  modify  in  any 
way  at  all  the  choice  that  has  been  made;  and  what  is  more,  the 
community  itself  cannot  control  so  much  as  a  single  word;  it  is 
bound  to  the  existing  language. 

No  longer  can  language  be  identified  with  a  contract  pure  and 
simple,  and  it  is  precisely  from  this  viewpoint  that  the  linguistic 
sign  is  a  particularly  interesting  object  of  study;  for  language 
furnishes  the  best  proof  that  a  law  accepted  by  a  community  is  a 
thing  that  is  tolerated  and  not  a  rule  to  which  all  freely  consent. 

Let  us  first  see  why  we  cannot  control  the  linguistic  sign  and  then 
draw  together  the  important  consequences  that  issue  from  the 

No  matter  what  period  we  choose  or  how  far  back  we  go,  lan- 
guage always  appears  as  a  heritage  of  the  preceding  period.  We 
might  conceive  of  an  act  by  which,  at  a  given  moment,  names  were 
assigned  to  things  and  a  contract  was  formed  between  concepts 
and  sound-images;  but  such  an  act  has  never  been  recorded.  The 
notion  that  things  might  have  happened  like  that  was  prompted 
by  our  acut£^a;feness-QLthe-^l5itFary_Jiature.  of  the  sign. 

No  society,  in  fact,  knows  or  has  ever  known  language  other  than 
as  a  product  inherited  from  preceding  generations,  and  one  to  be 
accepted  as  such.  That  is  why  the  question  of  the  origin  of  speech 


is  not  so  important  as  it  is  generally  assumed  to  be.  The  question^ 
is  not  even  worth  asking;  the  only  real  object  of  linguistics  is  the^ 
normal,  regular  life  of  an  existing-  idioin.,^A  particular  language-' 
state  is  always  the  product  of  historical  forces,  and  these  forces 
explain  why  the  sign  is  unchangeable,  i.e.  why  it  resists  any 
arbitrary  substitution. 

Nothing  is  explained  by  saying  that  language  is  something 
inherited  and  leaving  it  at  that.  Can  not  existing  and  inherited 
laws  be  modified  from  one  moment  to  the  next? 

To  meet  that  objection,  we  must  put  language  into  its  social 
setting  and  frame  the  question  just  as  we  would  for  any  other 
social  institution.  How  are  other  social  institutions  transmitted? 
This  more  general  question  includes  the  question  of  immutability. 
We  must  first  determine  the  greater  or  lesser  amounts  of  freedom 
that  the  other  institutions  enjoy;  in  each  instance  it  will  be  seen 
that  a  different  proportion  exists  between  fixed  tradition  and  the 
free  action  of  society.  The  next  step  is  to  discover  why  in  a  given 
category,  the  forces  of  the  first  type  carry  more  weight  or  less 
weight  than  those  of  the  second.  Finally,  coming  back  to  language, 
we  must  ask  why  the  historical  factor  of  transmission  dominates  it 
entirely  and  prohibits  any  sudden  widespread  change. 

There  are  many  possible  answers  to  the  question.  For  example, 
one  might  point  to  the  fact  that  succeeding  generations  are  not 
superimposed  on  one  another  like  the  drawers  of  a  piece  of  furni- 
ture, but  fuse  and  interpenetrate,  each  generation  embracing  in- 
dividuals of  all  ages— with  the  result  that  modifications  of  language 
are  not  tied  to  the  succession  of  generations.  One  might  also  recall 
the  sum  of  the  efforts  required  for  learning  the  mother  language 
and  conclude  that  a  general  change  would  be  impossible.  Again, 
it  might  be  added  that  reflection  does  not  enter  into  the  active  use 
of  an  idiom — speakers  are  largely  unconscious  of  the  laws  of  lan- 
guage; and  if  they  are  unaware  of  them,  how  could  they  modify 
them?  Even  if  they  were  aware  of  these  laws,  we  may  be  sure  that 
their  awareness  would  seldom  lead  to  criticism,  for  people  are 
generally  satisfied  with  the  language  they  have  received. 

The  foregoing  considerations  are  important  but  not  topical.  The 
following  are  more  basic  and  direct,  and  all  the  others  depend  on 


1)  The  arbitrary  nature  of  the  sign.  Above,  we  had  to  accept  the 
theoretical  possibility  of  change;  further  reflection  suggests  that 
the  arbitrary  nature  of  the  sign  is  really  what  protects  language 
from  any  attempt  to  modify  it.  Even  if  people  were  more  conscious 
of  language  than  they  are,  they  would  still  not  know  how  to  discuss 
it.  The  reason  is  simply  that  any  subject  in  order  to  be  discussed 
must  have  a  reasonable  basis.  It  is  possible,  for  instance,  to  discuss 
whether  the  monogamous  form  of  marriage  is  more  reasonable. than 
the  polygamous  form  and  to  advance  arguments  to  support  either 

^side.  One  could  also  argue  about  a  system  of  symbols,  for  the  sym- 
\  bol  has  a  rational  relationship  with  the  thing  signified  (see  p.  68) ; 
)  but  language  is  a  system  of  arbitrary  signs  and  lacks  the  necessary 
;  basis,  the  solid  ground  for  discussion.  There  is  no  reason  for 
-pfeferring  soeurjia^istexi^OchsAo  boeuf,  etc.       — "^ 

2)  The  multiplicity  of  signs  necessary  to  form  any  language. 
Another  important  deterrent  to  linguistic  change  is  the  great  num- 
ber of  signs  that  must  go  into  the  making  of  any  language.  A 
system  of  writing  comprising  twenty  to  forty  letters  can  in  case 
of  need  be  replaced  by  another  system.  The  same  would  be  true 
of  language  if  it  contained  a  limited  number  of  elements;  but 
linguistic  signs  are  numberless. 

3)  The  over-complexity  of  the  system.  A  language  constitutes  a 
system.  In  this  one  respect  (as  we  shall  see  later)  language  is  not 
completely  arbitrary  but  is  ruled  to  some  extent  by  logic;  it  is 
here  also,  however,  that  the  inability  of  the  masses  to  transform 
it  becomes  apparent.  The  system  is  a  complex  mechanism  that  can 
be  grasped  only  through  reflection ;  the  very  ones  who  use  it  daily 
are  ignorant  of  it.  We  can  conceive  of  a  change  only  through  the 
intervention  of  specialists,  grammarians,  logicians,  etc.;  but  ex- 
perience shows  us  that  all  such  meddlings  have  failed. 

4)  Collective  inertia  toward  innovation.  Language — and  this  con- 
sideration surpasses  all  the  others^ — is  at  every  moment  every- 
body's concern ;  spread  throughout  society  and  manipulated  by  it, 
language  is  something  used  daily  by  all.  Here  we  are  unable  to  set 
up  any  comparison  between  it  and  other  institutions.  The  pre- 
scriptions of  codes,  religious  rites,  nautical  signals,  etc.,  involve 
only  a  certain  number  of  individuals  simultaneously  and  then  only 


during  a  limited  period  of  time;  in  language,  on  the  contrary,  every- 
one participates  at  all  times,  and  that  is  why  it  is  constantly  being 
influenced  by  all.  This  capital  fact  suffices  to  show  the  impossibility 
of  revolution.  Of  all  social  institutions,  language  is  least  amenable 
to  initiative.  It  blends  with  the  life  of  society,  and  the  latter,  inert 
by  nature,  is  a  prime  conservative  force. 

But  to  say  that  language  is  a  product  of  social  forces  does  not 
suffice  to  show  clearly  that  it  is  unfree;  remembering  that  it  is 
always  the  heritage  ot  the  preceding  period,  we  must  add  that  these 
social  forces  are  linked  with  time.  Language  is  checked  not  only  by 
the  weight  of  the  collectivity  but  also  by  time.  These  two  are  in- 
separable. At  every  moment  solidarity  with  the  past  checks  free- 
dom of  choice.  We  say  man  and  dog.  This  does  not  prevent  the 
existence  in  the  total  phenomenon  of  a  bond  between  the  two 
antithetical  forces — arbitrary  convention  by  virtue  of  which  choice 
is  free  and  time  which  causes  choice  to  be  fixed.  Because  the  sign 
is  arbitrary,  it  follows  no  law  other  than  that  of  tradition,  and 
because  it  is  based  on  tradition,  it  is  arbitrary. 

2.  Mutability 

Time,  which  insures  the  continuity  of  language,  wields  another 
influence  apparently  contradictory  to  the  first:  the  more  or  less 
rapid  change  of  linguistic  signs.  In  a  certain  sense,  therefore,  we 
can  speak  of  both  the  immutability  and  the  mutability  of  the  sign.' 

In  the  last  analysis,  the  two  facts  are  interdependent:  the  sign 
is  exposed  to  alteration  because  it  perpetuates  itself.  What  pre- 
dominates in  all  change  is  the  persistence  of  the  old  substance; 
disregard  for  the  past  is  only  relative.  That  is  why  the  principle 
of  change  is  based  on  the  principle  of  continuity. 

Change  in  time  takes  many  forms,  on  any  one  of  which  an  im- 
portant chapter  in  linguistics  might  be  written.  Without  entering 
into  detail,  let  us  see  what  things  need  to  be  delineated. 

First,  let  there  be  no  mistake  about  the  meaning  that  we  attach 

to  the  word  change.  One  might  think  that  it  deals  especially  with 

^  It  would  be  wrong  to  reproach  F.  de  Saussure  for  being  illogical  or  para- 
doxical in  attributing  two  contradictory  qualities  to  language.  By  opposing 
two  striking  terms,  he  wanted  only  to  emphasize  the  fact  that  language  changes 
in  spite  of  the  inability  of  speakers  to  change  it.  One  can  also  say  that  it  is 
intangible  but  not  unchangeable.  [Ed.] 


phonetic  changes  undergone  by  the  signifier,  or  perhaps  changes  in 
meaning  which  affect  the  signified  concept.  That  view  would  be 
inadequate.  Regardless  of  what  the  forces  of  change  are,  whether 
in  isolation  or  in  combination,  they  always  result  in  a  shift  in  the 
relationship  between  the  signified  and  the  signifier. 

Here  are  some  examples.  Latin  necare  'kill'  became  noyer  'drown' 
in  French.  Both  the  sound-image  and  the  concept  changed;  but  it 
is  useless  to  separate  the  two  parts  of  the  phenomenon;  it  is 
sufficient  to  state  with  respect  to  the  whole  that  the  bond  between 
the  idea  and  the  sign  was  loosened,  and  that  there  was  a  shift  in 
their  relationship.  If  instead  of  comparing  Classical  Latin  necare 
with  French  noyer,  we  contrast  the  former  term  with  necare  of 
Vulgar  Latin  of  the  fourth  or  fifth  century  meaning  'drown'  the 
case  is  a  little  different;  but  here  again,  although  there  is  no 
appreciable  change  in  the  signifier,  there  is  a  shift  in  the  relation- 
ship between  the  idea  and  the  sign.* 

Old  German  dritteil  'one-third'  became  Drittel  in  Modern  Ger- 
man. Here,  although  the  concept  remained  the  same,  the  relation- 
ship was  changed  in  two  ways :  the  signifier  was  changed  not  only 
in  its  material  aspect  but  also  in  its  grammatical  form ;  the  idea  of 
Teil  'part'  is  no  longer  implied;  Drittel  is  a  simple  word.  In  one  way 
or  another  there  is  always  a  shift  in  the  relationship. 

In  Anglo-Saxon  the  preliterary  form  fot  'foot'  remained  while  its 

plural  *f6ti  became  fet  (Modern  English  feet) .  Regardless  of  the 

other  changes  that  are  implied,  one  thing  is  certain:  there  was  a 

shift  in  their  relationship;  other  correspondences  between  the 

phonetic  substance  and  the  idea  emerged.  -- — ^ 

I   Language  is  radically  powerless  to  defend  itself  against  the 

[forces  which  from  one  moment  to  the  next  are  shifting  the  relation- 

i  ship  between  the  signified  and  the  signifier.  This  is  one  of  the 

t consequences  of  the  arbitrary  nature  of  the  sign.  I 

Unlike  laligtra^e,  other  human  institutions — customs,  laws,  etc. 
— are  all  based  in  varying  degrees  on  the  natural  relations  of  things ; 
all  have  of  necessity  adapted  the  means  employed  to  the  ends 
pursued.  Even  fashion  in  dress  is  not  entirely  arbitrary;  we  can 
deviate  only  slightly  from  the  conditions  dictated  by  the  human 

*  From  May  to  July  of  1911,  Saussure  used  interchangeably  the  old  termi- 
nology {idea  and  sign)  and  the  new  {signified  and  signifier).  [Tr.] 


body.  Language  is  limited  by  nothing  in  the  choice  of  means,  for 
apparently  nothing  would  prevent  the  associating  of  any  idea 
whatsoever  with  just  any  sequence  of  sounds. 

To  emphasize  the  fact  that  language  is  a  genuine  institution, 
Whitney  quite  justly  insisted  upon  the  arbitrary  nature  of  signs; 
and  by  so  doing,  he  placed  linguistics  on  its  true  axis.  But  he  did 
not  follow  through  and  see  that  the  arbitrariness  of  language  radi- 
cally separates  it  from  all  other  institutions.  This  is  apparent  from 
the  way  in  which  language  evolves.  Nothing  could  be  more  com- 
plex. As  it  is  a  product  of  both  the  social  force  and  time,  no  one 
can  change  anything  in  it,  and  on  the  other  hand,  the  arbitrariness! 
of  its  signs  theoretically  entails  the  freedom  of  establishing_ju§l\ 
any  relationship  between  phonetic  substance  and  ideas.  iThe  result 
is  that  each  of  the  two  elements  united  in  the  sign  maintains  its 
own  life  to  a  degree  unknown  elsewhere,  and  that  language 
changes,  or  rather  evolves,  under  the  influence  of  all  the  forces 
which  can  affect  either  sounds  or  meanings.  The  evolution  is  in- 
evitable; there  is  no  example  of  a  single  language  that  resists  it. 
After  a  certain  period  of  time,  some  obvious  shifts  can  always  be 

Mutability  is  so  inescapable  that  it  even  holds  true  for  artificial 
languages.  Whoever  creates  a  language  controls  it  only  so  long  as 
it  is  not  in  circulation ;  from  the  moment  when  it  fulfills  its  mission 
and  becomes  the  property  of  everyone,  control  is  lost.  Take  Es- 
peranto as  an  example ;  if  it  succeeds,  will  it  escape  the  inexorable 
law?  Once  launched,  it  is  quite  likely  that  Esperanto  will  enter 
upon  a  fully  semiological  life;  it  will  be  transmitted  according  to 
laws  which  have  nothing  in  common  with  those  of  its  logical  cre- 
ation, and  there  will  be  no  turning  backwards.  A  man  proposing 
a  fixed  language  that  posterity  would  have  to  accept  for  what  it  is 
would  be  hke  a  hen  hatching  a  duck's  egg:  the  language  created 
by  him  would  be  borne  along,  willy-nilly,  by  the  current  that 
engulfs  all  languages. 

Signs  are  governed  by  a  principle  of  general  semiology:  con- 
tinuity in  time  is  coupled  to  change  in  time ;  this  is  confirmed  by 
orthographic  systems,  the  speech  of  deaf-mutes,  etc. 

But  what  supports  the  necessity  for  change?  I  might  be  re- 
proached for  not  having  been  as  explicit  on  this  point  as  on  the 
principle  of  immutability.  This  is  because  I  failed  to  distinguish 



between  the  different  forces  of  change.  We  must  consider  their 
great  variety  in  order  to  understand  the  extent  to  which  they  are 

The  causes  of  continuity  are  a  priori  within  the  scope  of  the 
observer,  but  the  causes  of  change  in  time  are  not.  It  is  better  not 
to  attempt  giving  an  exact  account  at  this  point,  but  to  restrict 
discussion  to  the  shifting  of  relationships  in  general.  Time  changes 
all  things;  there  is  no  reason  why  language  should  escape  this 
universal  law. 

Let  us  review  the  main  points  of  our  discussion  and  relate  them 
to  the  principles  set  up  in  the  Introduction. 

1)  Avoiding  sterile  word  definitions,  within  the  total  phenome- 
non represented  by  speech  we  first  singled  out  two  parts :  language 
and  speaking.  Language  is  speech  less  speaking.  It  is  the  whole  set 
of  linguistic  habits  which  allow  an  individual  to  understand  and 
to  be  understood. 

2)  But  this  definition  still  leaves  language  outside  its  social  con- 
text; it  makes  language  something  artificial  since  it  includes  only 
the  individual  part  of  reality;  for  the  realization  of  language,  a 
community  of  speakers  [masse  parlante]  is  necessary.  Contrary  to 
all  appearances,  language  never  exists  apart  from  the  social  fact, 
for  it  is  a  semiological  phenomenon.  Its  social  nature  is  one  of  its 
inner  characteristics.  Its  complete  definition  confronts  us  with  two 
inseparable  entities,  as  shown  in  this  drawing: 

But  under  the  conditions  described  language  is  not  living — it 
has  only  potential  life ;  we  have  considered  only  the  social,  not  the 
historical,  fact. 



3)  The  linguistic  sign  is  arbitrary;  language,  as  defined,  would 
therefore  seem  to  be  a  system  which,  because  it  depends  solely  on  a 
rational  principle,  is  free  and  can  be  organized  at  will.  Its  social 
nature,  considered  independently,  does  not  definitely  rule  out  this 
viewpoint.  Doubtless  it  is  not  on  a  purely  logical  basis  that  group 
psychology  operates;  one  must  consider  everything  that  deflects 
reason  in  actual  contacts  between  individuals.  But  the  thing  which 
keeps  language  from  being  a  simple  convention  that  can  be  modi- 
fied at  the  whim  of  interested  parties  is  not  its  social  nature ;  it  is 
rather  the  action  of  time  combined  with  the  social  force.  If  time 
is  left  out,  the  linguistic  facts  are  incomplete  and  no  conclusion 
is  possible. 

If  we  considered  language  in  time,  without  the  community  of 
speakers — imagine  an  isolated  individual  Uving  for  several  cen- 
turies— we  probably  would  notice  no  change;  time  would  not 
influence  language.  Conversely,  if  we  considered  the  community 
of  speakers  without  considering  time,  we  would  not  see  the  effect 
of  the  social  forces  that  influence  language.  To  represent  the  actual 
facts,  we  must  then  add  to  our  first  drawing  a  sign  to  indicate 
passage  of  time: 


Language  is  no  longer  free,  for  time  will  allow  the  social  forces 
at  work  on  it  to  carry  out  their  effects.  This  brings  us  back  to  the 
principle  of  continuity,  which  cancels  freedom.  But  continuity 
necessarily  implies  change,  varying  degrees  of  shifts  in  the  relation- 
ship between  the  signified  and  the  signifier. 


Chapter  III 

1.  Inner  Duality  of  All  Sciences  Concerned  with  Values 

Very  few  linguists  suspect  that  the  intervention  of  the  factor  of 
time  creates  difficulties  peculiar  to  linguistics  and  opens  to  their 
science  two  completely  divergent  paths. 

Most  other  sciences  are  unaffected  by  this  radical  duality ;  time 
produces  no  special  effects  in  them.  Astronomy  has  found  that  the 
stars  undergo  considerable  changes  but  has  not  been  obliged  on 
this  account  to  split  itself  into  two  disciplines.  Geology  is  con- 
cerned with  successions  at  almost  every  instant,  but  its  study  of 
strata  does  not  thereby  become  a  radically  distinct  discipline.  Law 
has  its  descriptive  science  and  its  historical  science;  no  one  opposes 
one  to  the  other.  The  political  history  of  states  is  unfolded  solely 
in  time,  but  a  historian  depicting  a  particular  period  does  not  work 
apart  from  history.  Conversely,  the  science  of  poHtical  institutions 
is  essentially  descriptive,  but  if  the  need  arises  it  can  easily  deal 
with  a  historical  question  without  disturbing  its  unity. 

On  the  contrary,  that  duality  is  already  forcing  itself  upon  the 
economic  sciences.  Here,  in  contrast  to  the  other  sciences,  political 
economy  and  economic  history  constitute  two  clearly  separated 
disciplines  within  a  single  science;  the  works  that  have  recently 
appeared  on  these  subjects  point  up  the  distinction.  Proceeding  as 
they  have,  economists  are — without  being  well  aware  of  it — 
obeying  an  inner  necessity.  A  similar  necessity  obliges  us  to  divide 
linguistics  into  two  parts,  each  with  its  own  principle.  Here  as  in 
political  economy  we  are  confronted  with  the  notion  of  value;  both 
sciences  are  concerned  with  a  system  for  equating  things  of  different 
orders — labor  and  wages  in  one  and  a  signified  and  signifier  in  the 

Certainly  all  sciences  would  profit  by  indicating  more  precisely 
the  co-ordinates  along  which  their  subject  matter  is  aligned.  Every- 



where  distinctions  should  be  made,  according  to  the  following 
illustration,  between  (1)  the  axis  of  simultaneities  (AB),  which 
stands  for  the  relations  of  coexisting  things  and  from  which  the 
intervention  of  time  is  excluded;  and  (2)  the  axis  of  successions 
(CD),  on  which  only  one  thing  can  be  considered  at  a  time  but 
upon  which  are  located  all  the  things  on  the  first  axis  together 
with  their  changes. 




For  a  science  concerned  with  values  the  distinction  is  a  practical 
necessity  and  sometimes  an  absolute  one.  In  these  fields  scholars 
cannot  organize  their  research  rigorously  without  considering  both 
co-ordinates  and  making  a  distinction  between  the  system  of 
values  per  se  and  the  same  values  as  they  relate  to  time. 

This  distinction  has  to  be  heeded  by  the  linguist  above  all  others, 
for  language  is  a  system  of  pure  values  which  are  determined  by 
nothing  except  the  momentary  arrangement  of  its  terms.  A  value 
— so  long  as  it  is  somehow  rooted  in  things  and  in  their  natural 
relations,  as  happens  with  economics  (the  value  of  a  plot  of  ground, 
for  instance,  is  related  to  its  productivity) — can  to  some  extent  be 
traced  in  time  if  we  remember  that  it  depends  at  each  moment 
upon  a  system  of  coexisting  values.  Its  link  with  things  gives  it, 
perforce,  a  natural  basis,  and  the  judgments  that  we  base  on  such 
values  are  therefore  never  completely  arbitrary;  their  variability 
is  limited.  But  we  have  just  seen  that  natural  data  have  no  place 
in  linguistics. 


Again,  the  more  complex  and  rigorously  organized  a  system  of 
values  is,  the  more  it  is  necessary,  because  of  its  very  complexity, 
to  study  it  according  to  both  co-ordinates.  No  other  system  em- 
bodies this  feature  to  the  same  extent  as  language.  Nowhere  else 
do  we  find  such  precise  values  at  stake  and  such  a  great  number 
and  diversity  of  terms,  all  so  rigidly  interdependent.  The  multi- 
plicity of  signs,  which  we  have  already  used  to  explain  the  con- 
tinuity of  language,  makes  it  absolutely  impossible  to  study 
simultaneously  relations  in  time  and  relations  within  the  system. 

The  reasons  for  distinguishing  two  sciences  of  language  are  clear. 
How  should  the  sciences  be  designated?  Available  terms  do  not  all 
bring  out  the  distinction  with  equal  sharpness.  "Linguistic  history" 
and  "historical  linguistics"  are  too  vague.  Since  political  history 
includes  the  description  of  different  periods  as  well  as  the  narration 
of  events,  the  student  might  think  that  he  is  studying  a  language 
according  to  the  axis  of  time  when  he  describes  its  successive  states, 
but  this  would  require  a  separate  study  of  the  phenomena  that 
make  language  pass  from  one  state  to  another.  Evolution  and 
evolutionary  linguistics  are  more  precise,  and  I  shall  use  these  ex- 
pressions often;  in  contrast,  we  can  speak  of  the  science  of  lan- 
guage-states [etats  de  langue]  or  static  linguistics. 

But  to  indicate  more  clearly  the  opposition  and  crossing  of  two 
orders  of  phenomena  that  relate  to  the  same  object,  I  prefer  to 
speak  of  synchronic  and  diachronic  linguistics.  Everything  that 
relates  to  the  static  side  of  our  science  is  synchronic;  everything 
that  has  to  do  with  evolution  is  diachronic.  Similarly,  synchrony 
and  diachrony  designate  respectively  a  language-state  and  an 
evolutionary  phase. 

2.  Inner  Duality  and  the  History  of  Linguistics 

The  first  thing  that  strikes  us  when  we  study  the  facts  of  lan- 
guage is  that  their  succession  in  time  does  not  exist  insofar  as  the 
speaker  is  concerned.  He  is  confronted  with  a  state.  That  is  why 
the  linguist  who  wishes  to  understand  a  state  must  discard  all 
knowledge  of  everything  that  produced  it  and  ignore  diachrony. 
He  can  enter  the  mind  of  speakers  only  by  completely  suppressing 
the  past.  The  intervention  of  history  can  only  falsify  his  judgment. 
It  would  be  absurd  to  attempt  to  sketch  a  panorama  of  the  Alps 


by  viewing  them  simultaneously  from  several  peaks  of  the  Jura; 
a  panorama  must  be  made  from  a  single  vantage  point.  The  same 
applies  to  language;  the  linguist  can  neither  describe  it  nor  draw 
up  standards  of  usage  except  by  concentrating  on  one  state.  When 
he  follows  the  evolution  of  the  language,  he  resembles  the  moving 
observer  who  goes  from  one  peak  of  the  Jura  to  another  in  order 
to  record  the  shifts  in  perspective. 

Ever  since  modern  linguistics  came  into  existence,  it  has  been 
completely  absorbed  in  diachrony.  Comparative  Indo-European 
philology  uses  the  materials  at  hand  to  reconstruct  hypothetically 
an  older  type  of  language;  comparison  is  but  a  means  of  recon- 
structing the  past.  The  method  is  the  same  in  the  narrower  study  of 
subgroups  (Romance  languages,  Germanic  languages,  etc.);  states 
intervene  only  irregularly  and  piecemeal.  Such  is  the  tendency 
introduced  by  Bopp.  His  conception  of  language  is  therefore  hybrid 
and  hesitating. 

Against  this,  what  was  the  procedure  of  those  who  studied  lan- 
guage before  the  beginning  of  modern  linguistics,  i.e.  the  "gram- 
marians" inspired  by  traditional  methods?  It  is  curious  to  note  that 
here  their  viewpoint  was  absolutely  above  reproach.  Their  works 
clearly  show  that  they  tried  to  describe  language-states.  Their 
program  was  strictly  synchronic.  The  Port  Royal  Grammar,  for 
example,  attempts  to  describe  the  state  of  French  under  Louis  XIV 
and  to  determine  its  values.  For  this,  the  language  of  the  Middle 
Ages  is  not  needed;  the  horizontal  axis  is  followed  faithfully  (see 
p.  80),  without  digression.  The  method  was  then  correct,  but  this 
does  not  mean  that  its  application  was  perfect.  Traditional  gram- 
mar neglects  whole  parts  of  language,  such  as  word  formation;  it 
is  normative  and  assumes  the  role  of  prescribing  rules,  not  of 
recording  facts ;  it  lacks  overall  perspective ;  often  it  is  unable  even 
to  separate  the  written  from  the  spoken  word,  etc. 

Classical  grammar  has  been  criticized  as  unscientific;  stiU,  its 
basis  is  less  open  to  criticism  and  its  data  are  better  defined  than 
is  true  of  the  linguistics  started  by  Bopp.  The  latter,  occupying 
ill-defined  ground,  has  no  clear-cut  objective.  It  straddles  two 
areas  because  it  is  unable  to  make  a  sharp  distinction  between 
states  and  successions. 

Linguistics,  having  accorded  too  large  a  place  to  history,  will 


turn  back  to  the  static  viewpoint  of  traditional  grammar  but 
in  a  new  spirit  and  with  other  procedures,  and  the  historical 
method  will  have  contributed  to  this  rejuvenation;  the  historical 
method  will  in  turn  give  a  better  understanding  of  language-states. 
The  old  grammar  saw  only  the  synchronic  fact;  linguistics  has 
revealed  a  new  class  of  phenomena;  but  that  is  not  enough;  one 
must  sense  the  opposition  between  the  two  classes  of  facts  to  draw 
out  all  its  consequences. 

3.  Inner  Duality  Illustrated  by  Examples 

The  opposition  between  the  two  viewpoints,  the  synchronic  and 
the  diachronic,  is  absolute  and  allows  no  compromise.  A  few  facts 
will  show  what  the  difference  is  and  why  it  is  irreducible. 

Latin  crispus  'crisp'  provided  French  with  the  root  crep-  from 
which  were  formed  the  verbs  crepir  'rough-cast'  and  decrepir 
'remove  mortar.'  Against  this,  at  a  certain  moment  the  word 
decrepitus,  of  unknown  origin,  was  borrowed  from  Latin  and  be- 
came decrepit  'decrepit.'  Certainly  today  the  community  of 
speakers  sets  up  a  relation  between  un  niur  decrypt  'a  wall  from 
which  mortar  is  falling'  and  U7i  homme  decrepit  'a  decrepit  man,' 
although  historically  the  two  words  have  nothing  in  common; 
people  often  speak  of  the  faqade  decrepite  of  a  house.  And  this  is 
static,  for  it  concerns  the  relation  between  two  coexisting  forms  of 
language.  For  its  realization,  the  concurrence  of  certain  evolu- 
tionary events  was  necessary.  The  pronunciation  of  crisp-  had  to 
become  crep-,  and  at  a  particular  moment  a  new  word  had  to  be 
borrowed  from  Latin.  It  is  obvious  that  the  diachronic  facts  are 
not  related  to  the  static  facts  which  they  produced.  They  belong 
to  a  different  class. 

Here  is  a  more  telhng  example.  In  Old  High  German  the  plural 
of  gast  'guest'  was  first  gasii,  that  of  hant  'hand'  was  hanti,  etc. 
Later  the  final  -i  produced  an  umlaut,  i.e.  it  resulted  in  the  chang- 
ing of  the  a  of  the  preceding  syllable  to  e:  gasti  —^  gesti;  hanti  -^ 
henti.  Then  the  final  -i  lost  its  timbre:  gesti  — >  geste,  etc.  The  result 
is  that  today  German  has  Gast:  Gdste,  Hand:  Hdnde,  and  a  whole 
group  of  words  marked  by  the  same  difference  between  the  singular 
and  the  plural.  A  very  similar  fact  occurred  in  Anglo-Saxon:  the 
earlier  forms  werefot:  *fdti,  top:  *tdH,  gos:  *gdsi,  etc.  Through  an 


initial  phonetic  change,  umlaut,  *fdti  became  *feti;  through  a  sec- 
ond, the  fall  of  final  -i,  feti  became  fet;  after  that,  fot  had  as  its 
plural  fet;  td]>,  te\>;  gos,  ges,  etc.  (Modern  English  foot:  feet,  tooth: 
teeth,  goose:  geese.) 

Previously,  when  speakers  used  gast:  gasti,  fot:  foti,  the  simple 
addition  of  an  i  marked  the  plural;  Gast:  Gaste  and  fot:  fet  show  a 
new  mechanism  for  indicating  the  plural.  The  mechanism  is  not 
the  same  in  both  instances;  in  Old  English  there  is  only  opposition 
between  vowels;  in  German  there  is  in  addition  the  presence  or 
absence  of  final  -e;  but  here  this  difference  is  unimportant. 

The  relation  between  a  singular  and  its  plural,  whatever  the 
forms  may  be,  can  be  expressed  at  each  moment  by  a  horizontal 

•  < >  •  Period  A 

•  < >  •  Period  B 

Whatever  facts  have  brought  about  passage  from  one  form  to 
another  should  be  placed  along  a  vertical  axis,  giving  the  overall 
picture : 

Period  A 

>  •   Period  B 

Our  illustration  suggests  several  pertinent  remarks: 

1)  In  no  way  do  diachronic  facts  aim  to  signal  a  value  by  means 
of  another  sign ;  that  gasti  became  gesti,  geste  {Gaste)  has  nothing  to 
do  with  the  plural  of  substantives ;  in  tragit  -^  tragi,  the  same  um- 
laut occurs  in  verbal  inflection,  and  so  forth.  A  diachronic  fact  is  an 
independent  event;  the  particular  synchronic  consequences  that 
may  stem  from  it  are  wholly  unrelated  to  it. 

2)  Diachronic  facts  are  not  even  directed  toward  changing  the 
system.  Speakers  did  not  wish  to  pass  from  one  system  of  relations 
to  another;  modification  does  not  affect  the  arrangement  but  rather 
its  elements. 

Here  we  again  find  the  principle  enunciated  previously:  never 
is  the  system  modified  directly.  In  itself  it  is  unchangeable;  only 
certain  elements  are  altered  without  regard  for  the  solidarity  that 
binds  them  to  the  whole.  It  is  as  if  one  of  the  planets  that  revolve 


around  the  sun  changed  its  dimensions  and  weight:  this  isolated 
event  would  entail  general  consequences  and  would  throw  the 
whole  system  out  of  equilibrium.  The  opposition  of  two  terms  is 
needed  to  express  plurality:  either  fot:  foti  or  fot:  Jet;  both  pro- 
cedures are  possible,  but  speakers  passed  from  one  to  the  other,  so 
to  speak,  without  having  a  hand  in  it.  Neither  was  the  whole  re- 
placed nor  did  one  system  engender  another;  one  element  in  the 
first  system  was  changed,  and  this  change  was  enough  to  give  rise 
to  another  system. 

3)  The  foregoing  observation  points  up  the  ever  fortuitous  nature 
of  a  state.  In  contrast  to  the  false  notion  that  we  readily  fashion 
for  ourselves  about  it,  language  is  not  a  mechanism  created  and 
arranged  with  a  view  to  the  concepts  to  be  expressed.  We  see  on 
the  contrary  that  the  state  which  resulted  from  the  change  was  not 
destined  to  signal  the  meaning  with  which  it  was  impregnated.  In 
a  fortuitous  state  {fot:  fet),  speakers  took  advantage  of  an  exist- 
ing difference  and  made  it  signal  the  distinction  between  singu- 
lar and  plural;  fot:  fet  is  no  better  for  this  purpose  than  fot:  *foti. 
In  each  state  the  mind  infiltrated  a  given  substance  and  breathed 
life  into  it.  This  new  perspective,  inspired  by  historical  linguistics, 
is  unknown  to  traditional  grammar,  which  could  never  acquire  it 
by  its  own  methods.  Most  philosophers  of  language  are  equally 
ignorant  of  it,  and  yet  nothing  is  more  important  from  the  philo- 
sophical viewpoint. 

4)  Are  facts  of  the  diachronic  series  of  the  same  class,  at  least, 
as  facts  of  the  synchronic  series?  By  no  means,  for  we  have  seen 
that  changes  are  wholly  unintentional  while  the  synchronic  fact  is 
always  significant.  It  always  calls  forth  two  simultaneous  terms. 
Not  Gaste  alone  but  the  opposition  Gast:  Gdste  expresses  the  plural. 
The  diachronic  fact  is  just  the  opposite:  only  one  term  is  involved, 
and  for  the  new  one  to  appear  (Gdste),  the  old  one  (gasti)  must 
first  give  way  to  it. 

To  try  to  unite  such  dissimilar  facts  in  the  same  discipline  would 
certainly  be  a  fanciful  undertaking.  The  diachronic  perspective 
deals  with  phenomena  that  are  unrelated  to  systems  although  they 
do  condition  them. 

Here  are  some  other  examples  to  strengthen  and  complement  the 
conclusions  drawn  from  the  first  ones. 


In  French,  the  accent  always  falls  on  the  last  syllable  unless  this 
syllable  contains  a  mute  e  (o).  This  is  a  synchronic  fact,  a  relation 
between  the  whole  set  of  French  words  and  accent.  What  is  its 
source?  A  previous  state.  Latin  had  a  different  and  more  compli- 
cated system  of  accentuation:  the  accent  was  on  the  penultimate 
syllable  when  the  latter  was  long ;  when  short,  the  accent  fell  back 
on  the  antepenult  (cf.  amicus,  dnima).  The  Latin  law  suggests 
relations  that  are  in  no  way  analogous  to  the  French  law.  Doubtless 
the  accent  is  the  same  in  the  sense  that  it  remained  in  the  same 
position ;  in  French  words  it  always  falls  on  the  syllable  that  had  it 
in  Latin :  amtcum  — >  ami,  dnimum  -^  dme.  But  the  two  formulas 
are  different  for  the  two  moments  because  the  forms  of  the  words 
changed.  We  know  that  everything  after  the  accent  either  dis- 
appeared or  was  reduced  to  mute  e.  As  a  result  of  the  alteration  of 
the  word,  the  position  of  the  accent  with  respect  to  the  whole  was 
no  longer  the  same;  subsequently  speakers,  conscious  of  the  new 
relation,  instinctively  put  the  accent  on  the  last  syllable,  even  in 
borrowed  words  introduced  in  their  written  forms  (facile,  consul, 
ticket,  burgrave,  etc.).  Speakers  obviously  did  not  try  to  change 
systems,  to  apply  a  new  formula,  since  in  words  like  amtcum  — ^  ami 
the  accent  always  remained  on  the  same  syllable ;  but  a  diachronic 
fact  w^as  interposed:  speakers  changed  the  position  of  the  accent 
without  having  a  hand  in  it.  A  law  of  accentuation,  like  everything 
that  pertains  to  the  linguistic  system,  is  an  arrangement  of  terms, 
a  fortuitous  and  involuntary  result  of  evolution. 

Here  is  an  even  more  striking  example.  In  Old  Slavic,  slovo  'word' 
has  in  the  instrumental  singular  slovem'  b,  in  the  nominative  plural 
slova,  in  the  genitive  plural  slov'b,  etc.;  in  the  declension  each  case 
has  its  own  ending.  But  today  the  weak  vowels  b  and  'b,  Slavic 
representatives  of  Proto-Indo-European  i  and  m,  have  disappeared. 
Czech,  for  example,  has  slovo,  slovem,  slova,  slov;  Ukewise  zena 
'woman' :  accusative  singular  zenu,  nominative  plural  zeny,  genitive 
plural  zen.  Here  the  genitive  {slov,  zen)  has  zero  inflection.  We  see 
then  that  a  material  sign  is  not  necessary  for  the  expression  of  an 
idea;  language  is  satisfied  with  the  opposition  between  something 
and  nothing.  Czech  speakers  recognize  zen  as  a  genitive  plural 
simply  because  it  is  neither  zena  nor  zenu  nor  any  of  the  other 
forms.  It  seems  strange  at  first  glance  that  such  a  particular  notion 


as  that  of  the  genitive  plural  should  have  taken  the  zero  sign,  but 
this  very  fact  proves  that  everything  comes  about  through  sheer 
accident.  Language  is  a  mechanism  that  continues  to  function  in 
spite  of  the  deteriorations  to  which  it  is  subjected. 

All  this  confirms  the  principles  previously  stated.  To  summarize: 
Language  is  a  system  whose  parts  can  and  must  all  be  considered 
in  their  synchronic  solidarity. 

Since  changes  never  affect  the  system  as  a  whole  but  rather  one 
or  another  of  its  elements,  they  can  be  studied  only  outside  the 
system.  Each  alteration  doubtless  has  its  countereffect  on  the  sys- 
tem, but  the  initial  fact  affected  only  one  point;  there  is  no  inner 
bond  between  the  initial  fact  and  the  effect  that  it  may  subse- 
quently produce  on  the  whole  system.  The  basic  difference  between 
successive  terms  and  coexisting  terms,  between  partial  facts  and 
facts  that  affect  the  system,  precludes  making  both  classes  of  fact 
the  subject  matter  of  a  single  science. 

4.  The  Difference  between  the  Two  Classes  Illustrated  by  Comparisons 
To  show  both  the  autonomy  and  the  interdependence  of  syn- 
chrony we  can  compare  the  first  to  the  projection  of  an  object  on  a 
plane  surface.  Any  projection  depends  directly  on  the  nature  of  the 
object  projected,  yet  differs  from  it — the  object  itself  is  a  thing 
apart.  Otherwise  there  would  not  be  a  whole  science  of  projections; 
considering  the  bodies  themselves  would  suffice.  In  linguistics  there 
is  the  same  relationship  between  the  historical  facts  and  a  lan- 
guage-state, which  is  hke  a  projection  of  the  facts  at  a  particular 
moment.  We  do  not  learn  about  synchronic  states  by  studying 
bodies,  i.e.  diachronic  events,  any  more  than  we  learn  about  geo- 
metric projections  by  studying,  even  carefully,  the  different  types 
of  bodies. 

Similarly  if  the  stem  of  a  plant  is  cut  transversely,  a  rather  com- 
plicated design  is  formed  by  the  cut  surface ;  the  design  is  simply 
one  perspective  of  the  longitudinal  fibers,  and  we  would  be  able  to 
see  them  on  making  a  second  cut  perpendicular  to  the  first.  Here 
again  one  perspective  depends  on  the  other;  the  longitudinal  cut 
shows  the  fibers  that  constitute  the  plant,  and  the  transversal  cut 
shows  their  arrangement  on  a  particular  plane;  but  the  second  is 
distinct  from  the  first  because  it  brings  out  certain  relations  be- 



tween  the  fibers— relations  that  we  could  never  grasp  by  viewing 
the  longitudinal  plane. 

But  of  all  comparisons  that  might  be  imagined,  the  most  friutful 
is  the  one  that  might  be  drawn  between  the  functioning  of  language 
and  a  game  of  chess.  In  both  instances  we  are  confronted  with  a 
system  of  values  and  their  observable  modifications.  A  game  of 
chess  is  like  an  artificial  realization  of  what  language  offers  in  a 
natural  form. 

Let  us  examine  the  matter  more  carefully. 

First,  a  state  of  the  set  of  chessmen  corresponds  closely  to  a  state 
of  language.  The  respective  value  of  the  pieces  depends  on  their 
position  on  the  chessboard  just  as  each  linguistic  term  derives  its 
value  from  its  opposition  to  all  the  other  terms. 

In  the  second  place,  the  system  is  always  momentary;  it  varies 
from  one  position  to  the  next.  It  is  also  true  that  values  depend 
above  all  else  on  an  unchangeable  convention,  the  set  of  rules  that 
exists  before  a  game  begins  and  persists  after  each  move.  Rules  that 
are  agreed  upon  once  and  for  all  exist  in  language  too;  they  are  the 
constant  principles  of  semiology. 

Finally,  to  pass  from  one  state  of  equilibrium  to  the  next,  or — 
according  to  our  terminology — from  one  synchrony  to  the  next, 
only  one  chesspiece  has  to  be  moved ;  there  is  no  general  rummage. 
Here  we  have  the  counterpart  of  the  diachronic  phenomenon  with 
all  its  peculiarities.  In  fact : 

(a)  In  each  play  only  one  chesspiece  is  moved ;  in  the  same  way 
in  language,  changes  affect  only  isolated  elements. 


(b)  In  spite  of  that,  the  move  has  a  repercussion  on  the  whole 
system;  it  is  impossible  for  the  player  to  foresee  exactly  the  extent 
of  the  effect.  Resulting  changes  of  value  will  be,  according  to  the 
circumstances,  either  nil,  very  serious,  or  of  average  importance. 
A  certain  move  can  revolutionize  the  whole  game  and  even  affect 
pieces  that  are  not  immediately  involved.  We  have  just  seen  that 
exactly  the  same  holds  for  language. 

(c)  In  chess,  each  move  is  absolutely  distinct  from  the  preceding 
and  the  subsequent  equilibrium.  The  change  effected  belongs  to 
neither  state:  only  states  matter. 

In  a  game  of  chess  any  particular  position  has  the  unique  char- 
acteristic of  being  freed  from  all  antecedent  positions;  the  route 
used  in  arriving  there  makes  absolutely  no  difference;  one  who  has 
followed  the  entire  match  has  no  advantage  over  the  curious  party 
who  comes  up  at  a  critical  moment  to  inspect  the  state  of  the  game ; 
to  describe  this  arrangement,  it  is  perfectly  useless  to  recall  what 
had  just  happened  ten  seconds  previously.  All  this  is  equally  ap- 
pUcable  to  language  and  sharpens  the  radical  distinction  between 
diachrony  and  synchrony.  Speaking  operates  only  on  a  language- 
state,  and  the  changes  that  intervene  between  states  have  no  place 
in  either  state. 

At  only  one  point  is  the  comparison  weak:  the  chessplayer 
intends  to  bring  about  a  shift  and  thereby  to  exert  an  action  on  the 
system,  whereas  language  premeditates  nothing.  The  pieces  of  lan- 
guage are  shifted — or  rather  modified— spontaneously  and  for- 
tuitously. The  umlaut  of  Hdnde  for  hanti  and  Gdste  for  gasti  (see 
p.  83)  produced  a  new  system  for  forming  the  plural  but  also  gave 
rise  to  verbal  forms  hke  tragi  from  tragit,  etc.  In  order  to  make  the 
game  of  chess  seem  at  every  point  like  the  functioning  of  language, 
we  would  have  to  imagine  an  unconscious  or  unintelligent  player. 
This  sole  difference,  however,  makes  the  comparison  even  more 
instructive  by  showing  the  absolute  necessity  of  making  a  distinc- 
tion between  the  two  classes  of  phenomena  in  linguistics.  For  if 
diachronic  facts  cannot  be  reduced  to  the  synchronic  system  which 
they  condition  when  the  change  is  intentional,  all  the  more  will 
they  resist  when  they  set  a  blind  force  against  the  organization  of 
a  system  of  signs. 


5.  The  Two  Linguistics  Contrasted  According  to  Their  Methods  and 


Everywhere  the  opposition  between  diachrony  and  synchrony 
stands  out. 

For  instance — and  to  begin  with  the  most  apparent  fact — they 
are  not  of  equal  importance.  Here  it  is  evident  that  the  synchronic 
viewpoint  predominates,  for  it  is  the  true  and  only  reality  to  the 
community  of  speakers  (see  p.  81).  The  same  is  true  of  the  lin- 
guist: if  he  takes  the  diachronic  perspective,  he  no  longer  observes 
language  but  rather  a  series  of  events  that  modify  it.  People  often 
affirm  that  nothing  is  more  important  than  understanding  the 
genesis  of  a  particular  state;  this  is  true  in  a  certain  sense:  the 
forces  that  have  shaped  the  state  illuminate  its  true  nature,  and 
knowing  them  protects  us  against  certain  illusions  (see  pp.  84  ff.) ; 
but  this  only  goes  to  prove  clearly  that  diachronic  linguistics  is  not 
an  end  in  itself.  What  is  said  of  journalism  applies  to  diachrony: 
it  leads  everywhere  if  one  departs  from  it. 

The  methods  of  diachrony  and  synchrony  also  differ,  and  in  two 

(a)  Synchrony  has  only  one  perspective,  the  speakers',  and  its 
whole  method  consists  of  gathering  evidence  from  speakers;  to 
know  to  just  what  extent  a  thing  is  a  reality,  it  is  necessary  and 
sufficient  to  determine  to  what  extent  it  exists  in  the  minds  of 
speakers.  Diachronic  linguistics,  on  the  contrary,  must  distinguish 
two  perspectives.  One  of  these,  the  prospective,  follows  the  course 
of  time ;  the  other,  the  retrospective,  goes  back  in  time ;  the  result  is 
a  duphcation  in  methodology  with  which  we  shall  deal  in  Part  Five. 

(b)  A  second  difference  results  from  delimiting  the  fields  em- 
braced by  each  of  the  two  disciplines.  Synchronic  study  has  as  its 
object,  not  everything  that  is  simultaneous,  but  only  the  totahty 
of  facts  corresponding  to  each  language;  separation  will  go  as  far 
as  dialects  and  subdialects  when  necessary.  The  term  synchronic 
is  really  not  precise  enough;  it  should  be  replaced  by  another — 
rather  long  to  be  sure — idiostjnchronic.  Against  this,  diachronic 
linguistics  not  only  does  not  need  but  even  rejects  such  special- 
ization; the  terms  that  it  studies  do  not  necessarily  belong  to  the 
same  language  (compare  Proto-Indo-European  *esti,  Greek  esti, 


German  ist,  and  French  est).  The  succession  of  diachronic  events 
and  their  multiplication  in  space  are  precisely  what  creates  the 
diversity  of  idioms.  To  justify  the  associating  of  two  forms,  it  is 
enough  to  show  that  they  are  connected  by  a  historical  bond, 
however  indirect  it  may  be. 

The  foregoing  oppositions  are  neither  the  most  striking  nor  the 
most  profound.  One  consequence  of  the  radical  antimony  between 
the  evolutionary  and  the  static  fact  is  that  all  notions  associated 
with  one  or  the  other  are  to  the  same  extent  mutually  irreducible. 
Any  notion  will  point  up  this  truth.  The  synchronic  and  diachronic 
"phenomenon,"  for  example,  have  nothing  in  common  (see  p.  85). 
One  is  a  relation  between  simultaneous  elements,  the  other  the 
substitution  of  one  element  for  another  in  time,  an  event. 

We  shall  also  see  (p.  107)  that  diachronic  and  S3nichronic  identi- 
ties are  two  very  different  things ;  historically  the  French  negation 
pas  is  identical  to  the  substantive  pas  'step,'  whereas  the  two  forms 
are  distinct  in  modern  French.  These  observations  would  suffice  to 
show  the  necessity  of  not  confusing  the  two  viewpoints,  but  no- 
where is  this  necessity  more  apparent  than  in  the  distinction  we 
are  about  to  make. 

6.  Synchronic  and  Diachronic  Law 

It  is  a  popular  practice  to  speak  of  laws  in  linguistics.  But  are 
the  facts  of  language  actually  governed  by  laws?  If  so,  what  are 
they  like?  Since  language  is  a  social  institution,  one  might  assume 
a  priori  that  it  is  governed  by  prescriptions  analogous  to  those  that 
control  communities.  Now  every  social  law  has  two  basic  charac- 
teristics: it  is  imperative  and  it  is  general;  it  comes  in  by  force  and 
it  covers  all  cases — within  certain  limits  of  time  and  place,  of 

Do  the  laws  of  language  fit  this  definition?  The  first  step  in 
answering  the  question — in  line  with  what  has  just  been  said — is 
to  separate  once  more  the  synchronic  and  diachronic  areas.  The 
two  problems  must  not  be  confused;  speaking  of  linguistic  law  in 
general  is  like  trying  to  pin  down  a  ghost. 

Here  are  some  examples,  taken  from  Greek,  in  which  the  two 
classes  are  intentionally  jumbled: 


1.  Proto-Indo-European  voiced  aspirates  became  voiceless: 
*dhumos  — >  thumos  'breath  of  life,'  *hhero  —^  phero  'I  bear/  etc. 

2.  The  accent  never  falls  farther  back  than  the  antepenult. 

3.  All  words  end  in  a  vowel  or  in  s,  n,  or  r,  to  the  exclusion  of  all 
other  consonants. 

4.  Pre  vocalic  initial  s  became  h  (sign  of  aspiration) :  *septm 
(Latin  septem)  -^  heptd. 

5.  Final  m  changed  to  n:  *jugom  — *  zugon  (cf.  Latin  jugum).^ 

6.  Final  occlusives  fell:  *gunaik -^  gunai,  *epherst —^  ephere, 
*epheront  -^  epheron. 

Law  1  is  diachronic :  dh  became  th,  etc.  Law  2  expresses  a  relation 
between  the  word-unit  and  accent,  a  sort  of  contract  between  two 
coexisting  terms ;  it  is  a  synchronic  law.  The  same  is  true  of  Law  3 
since  it  concerns  the  word-unit  and  its  ending.  Laws  4,  5,  and  6  are 
diachronic:  s  became  h;  -n  replaced  -m;  -t,  -k,  etc.  disappeared 
without  leaving  a  trace. 

We  should  also  notice  that  Law  3  is  the  result  of  5  and  6;  two 
diachronic  facts  created  a  synchronic  fact. 

After  we  separate  the  two  classes  of  laws,  we  see  that  Laws  2  and 
3  are  basically  different  from  Laws  1,  4,  5,  and  6. 

The  synchronic  law  is  general  but  not  imperative.  Doubtless  it 
is  imposed  on  individuals  by  the  weight  of  collective  usage  (see 
p.  73),  but  here  I  do  not  have  in  mind  an  obhgation  on  the  part 
of  speakers.  I  mean  that  in  language  no  force  guarantees  the  main- 
tenance of  a  regularity  when  established  on  some  point.  Being  a 
simple  expression  of  an  existing  arrangement,  the  sjmchronic  law 
reports  a  state  of  affairs ;  it  is  like  a  law  that  states  that  trees  in  a 
certain  orchard  are  arranged  in  the  shape  of  a  quincunx.  And  the 
arrangement  that  the  law  defines  is  precarious  precisely  because 
it  is  not  imperative.  Nothing  is  more  regular  than  the  synchronic 
law  that  governs  Latin  accentuation  (a  law  comparable  in  every 
way  to  Law  2  above);  but  the  accentual  rule  did  not  resist  the 

^  According  to  Meillet  (Mem.  de  la  Soc.  de  Ldng.,  IX,  pp.  365  ff.)  and 
Gauthiot  {La  fin  du  mot  indo-europeen,  pp.  158  ff.),  final  -m  did  not  exist  in 
Proto-Indo-European,  which  used  only  -n;  if  this  theory  is  accepted,  Law  5 
can  be  stated  in  this  way:  Greek  preserved  every  final  -n;  its  demonstrative 
value  is  not  diminished  since  the  phonetic  phenomenon  that  results  in  the 
preservation  of  a  former  state  is  the  same  in  nature  as  the  one  that  manifests 
a  change  (see  p.  145).  [Ed.] 


forces  of  alteration  and  gave  way  to  a  new  law,  the  one  of  French 
(see  abo"ve  p.  86).  In  short,  if  one  speaks  of  law  in  synchrony, 
it  is  in  the  sense  of  an  arrangement,  a  principle  of  regularity. 

Diachrony,  on  the  contrary,  supposes  a  dynamic  force  through 
which  an  effect  is  produced,  a  thing  executed.  But  this  imperative- 
ness is  not  sufficient  to  warrant  applying  the  concept  of  law  to 
evolutionary  facts;  we  can  speak  of  law  only  when  a  set  of  facts 
obeys  the  same  rule,  and  in  spite  of  certain  appearances  to  the 
contrary,  diachronic  events  are  always  accidental  and  particular. 

The  accidental  and  particular  character  of  semantic  facts  is  im- 
mediately apparent.  That  French  poutre  'mare'  has  acquired  the 
meaning  'piece  of  wood,  rafter'  is  due  to  particular  causes  and  does 
not  depend  on  other  changes  that  might  have  occurred  at  the  same 
time.  It  is  only  one  accident  among  all  those  registered  in  the 
history  of  the  language. 

As  for  syntactical  and  morphological  transformations,  the  issue 
is  not  so  clear  from  the  outset.  At  a  certain  time  almost  all  old 
subject-case  forms  disappeared  in  French.  Here  a  set  of  facts  ap- 
parently obeys  the  same  law.  But  such  is  not  the  case,  for  all  the 
facts  are  but  multiple  manifestations  of  one  and  the  same  isolated 
fact.  The  particular  notion  of  subject  was  affected,  and  its  dis- 
appearance naturally  caused  a  whole  series  of  forms  to  vanish.  For 
one  who  sees  only  the  external  features  of  language,  the  unique 
phenomenon  is  drowned  in  the  multitude  of  its  manifestations. 
Basically,  however,  there  is  but  one  phenomenon,  and  this  histori- 
cal event  is  just  as  isolated  in  its  own  order  as  the  semantic  change 
undergone  by  poutre.  It  takes  on  the  appearance  of  a  "law"  only 
because  it  is  realized  within  a  system.  The  rigid  arrangement  of  the 
system  creates  the  illusion  that  the  diachronic  fact  obeys  the  same 
rules  as  the  synchronic  fact. 

Finally,  as  regards  phonetic  changes,  exactly  the  same  is  true. 
Yet  the  popular  practice  is  to  speak  of  phonetic  laws.  Indeed,  it  is 
said  that  at  a  given  time  and  in  a  given  area  all  words  having 
the  same  phonic  features  are  affected  by  the  same  change;  for 
example.  Law  1  on  page  92  {*dhumos  —^  Greek  thumos)  affects  all 
Greek  words  containing  a  voiced  aspirate  (cf .  *nebhos  — »  nephos, 
*medhu  — ^  methu,  *anghd  -^  dnkho,  etc.) ;  Law  4  {*septm  -^  heptd) 
applies  to  *serpd  -^  herpo,  *sus  —*■  hUs,  and  to  all  words  that  begin 


with  s.  This  regularity,  which  has  at  times  been  disputed,  is  ap- 
parently firmly  established;  obvious  exceptions  do  not  lessen  the 
inevitability  of  such  changes,  for  they  can  be  explained  either  by 
more  special  phonetic  laws  (see  the  example  of  trikhes:  thriksi, 
p.  97)  or  by  the  interference  of  facts  of  another  class  (analogy,  etc.). 
Nothing  seems  to  fit  better  the  definition  given  above  for  the 
word  law.  And  yet,  regardless  of  the  number  of  instances  where  a 
phonetic  law  holds,  all  facts  embraced  by  it  are  but  manifestations 
of  a  single  particular  fact. 

The  real  issue  is  to  find  out  whether  phonetic  changes  affect 
v;ords  or  only  sounds,  and  there  is  no  doubt  about  the  answer:  in 
nephos,  methu,  ankho,  etc.  a  certain  phoneme — a  voiced  Proto- 
Indo-European  aspirate — became  voiceless,  Proto-Greek  initial  s 
became  h,  etc.;  each  fact  is  isolated,  independent  of  the  other 
events  of  the  same  class,  independent  also  of  the  words  in  which 
the  change  took  place.®  The  phonic  substance  of  all  the  words  was 
of  course  modified,  but  this  should  not  deceive  us  as  to  the  real 
nature  of  the  phenomenon. 

What  supports  the  statement  that  words  themselves  are  not 

directly  involved  in  phonetic  transformations?  The  very  simple 

observation  that  these  transformations  are  basically  alien  to  words 

and  cannot  touch  their  essence.  The  word-unit  is  not  constituted 

solely  by  the  totahty  of  its  phonemes  but  by  characteristics 

other  than  its  material  quality.  Suppose  that  one  string  of  a  piano 

is  out  of  tune:  a  discordant  note  will  be  heard  each  time  the  one 

who  is  playing  a  melody  strikes  the  corresponding  key.  But  where 

is  the  discord?  In  the  melody?  Certainly  not;  the  melody  has  not 

been  affected;  only  the  piano  has  been  impaired.  Exactly  the  same 

is  true  in  phonetics.  Our  system  of  phonemes  is  the  instrument  we 

play  in  order  to  articulate  the  words  of  language;  if  one  of  its 

elements  is  modified,  diverse  consequences  may  ensue,  but  the 

modification  itself  is  not  concerned  with  the  words  which  are,  in 

a  manner  of  speaking,  the  melodies  of  our  repertory. 

^  Of  course  the  examples  cited  above  are  purely  schematic :  linguistics  is 
right  in  trying  currently  to  relate  to  the  same  initial  principle  the  largest 
possible  series  of  phonetic  changes;  for  instance,  Meillet  explains  all  the 
transformations  of  Greek  occlusives  by  progressive  weakening  of  their  articu- 
lation (see  Mem.  de  la  Soc.  de  Ling.,  IX,  pp.  163  ff.).  Naturally  the  conclusions 
on  the  nature  of  phonetic  changes  are  in  the  last  analysis  apphcable  to  these 
general  facts,  wherever  they  exist.  [Ed.] 


Diachronic  facts  are  then  particular;  a  shift  in  a  system  is 
brought  about  by  events  which  not  only  are  outside  the  system 
(see  p.  84),  but  are  isolated  and  form  no  system  among  them- 

To  summarize:  synchronic  facts,  no  matter  what  they  are, 
evidence  a  certain  regularity  but  are  in  no  way  imperative;  dia- 
chronic facts,  on  the  contrary,  force  themselves  upon  language 
but  are  in  no  way  general. 

In  a  word — and  this  is  the  point  I  have  been  trying  to  make — 
neither  of  the  two  classes  of  facts  is  governed  by  laws  in  the  sense 
defined  above,  and  if  one  still  wishes  to  speak  of  linguistic  laws,  the 
word  will  embrace  completely  different  meanings,  depending  on 
whether  it  designates  facts  of  one  class  or  the  other. 

7.  7s  There  a  Panchronic  Viewpoint? 

Up  to  this  point  the  term  law  has  been  used  in  the  legal  sense. 
But  cannot  the  term  also  be  used  in  language  as  in  the  physical  and 
natural  sciences,  i.e.  in  the  sense  of  relations  that  are  everywhere 
and  forever  verifiable?  In  a  word,  can  not  language  be  studied 
from  a  panchronic  viewpoint? 

Doubtless.  Since  phonetic  changes  have  always  occurred  and 
are  still  occurring,  this  general  phenomenon  is  a  permanent  char- 
acteristic of  speech;  it  is  therefore  one  of  the  laws  of  speech.  In 
linguistics  as  in  chess  (see  pp.  88  ff.)  there  are  rules  that  outlive 
all  events.  But  they  are  general  principles  existing  independently 
of  concrete  facts.  When  we  speak  of  particular,  tangible  facts, 
there  is  no  panchronic  viewpoint.  Each  phonetic  change,  regardless 
of  its  actual  spread,  is  Hmited  to  a  definite  time  and  territory;  no 
change  occurs  at  all  times  and  in  all  places;  change  exists  only 
diachronically.  These  general  principles  are  precisely  what  serve 
as  a  criterion  for  determining  what  belongs  to  language  and  what 
does  not.  A  concrete  fact  that  lends  itself  to  panchronic  explanation 
cannot  belong  to  language.  Take  the  French  word  chose  'thing': 
from  the  diachronic  viewpoint  it  stands  in  opposition  to  the  Latin 
word  from  which  it  derives,  causa;  from  the  synchronic  viewpoint 
it  stands  in  opposition  to  every  word  that  might  be  associated  with 
it  in  Modern  French.  Only  the  sounds  of  the  word  considered  in- 
dependently {§oz)  are  susceptible  of  panchronic  observation,  but 


they  have  no  linguistic  value.  Even  from  the  panchronic  viewpoint 
§gz,  considered  in  a  chain  like  iin  §oz  odmirahld  'an  admirable  thing/ 
is  not  a  unit  but  a  shapeless  mass;  indeed,  why  ^oz  rather  than  oza 
or  nsof  It  is  not  a  value,  for  it  has  no  meaning.  From  the  pan- 
chronic  viewpoint  the  particular  facts  of  language  are  never 

8.  Consequences  of  the  Confusing  of  Synchrony  and  Diachrony 
Two  instances  will  be  cited : 

(a)  Synchronic  truth  seems  to  be  the  denial  of  diachronic  truth, 
and  one  who  has  a  superficial  view  of  things  imagines  that  a  choice 
must  be  made;  this  is  really  unnecessary;  one  truth  does  not  ex- 
clude the  other.  That  French  deyit  'spite'  originally  meant  con- 
tempt does  not  prevent  the  word  from  having  a  completely 
different  meaning  now;  etymology  and  synchronic  value  are  dis- 
tinct. Similarly,  traditional  grammar  teaches  that  the  present 
participle  is  variable  and  shows  agreement  in  the  same  manner  as 
an  adjective  in  certain  cases  in  Modern  French  (cf.  une  eau 
courante  'running  water')  but  is  invariable  in  others  (cf.  une  per- 
sonne  courant  dans  la  rue  'a  person  running  in  the  street').  But 
historical  grammar  shows  that  it  is  not  a  question  of  one  and  the 
same  form :  the  first  is  the  continuation  of  the  variable  Latin  par- 
ticiple (currentum)  while  the  second  comes  from  the  invariable 
ablative  form  of  the  gerund  {currendo)?  Does  synchronic  truth 
contradict  diachronic  truth,  and  must  one  condemn  traditional 
granmiar  in  the  name  of  historical  grammar?  No,  for  that  would  be 
seeing  only  half  of  the  facts;  one  must  not  think  that  the  historical 
fact  alone  matters  and  is  sufficient  to  constitute  language.  Doubt- 
less from  the  viewpoint  of  its  origin  the  participle  courant  has  two 
elements,  but  in  the  collective  mind  of  the  community  of  speakers, 
these  are  drawn  together  and  fused  into  one.  The  synchronic  truth 
is  just  as  absolute  and  indisputable  as  the  diachronic  truth. 

(b)  Synchronic  truth  is  so  similar  to  diachronic  truth  that  people 

confuse  the  two  or  think  it  superfluous  to  separate  them.  For 

example,  they  try  to  explain  the  meaning  of  French  pere  'father' 

^  This  generally  accepted  theory  has  been  recently  but,  we  believe,  un- 
successfully attacked  by  M.  E.  Larch  {Das  invariable  Participium  praesentis, 
Erlangen,  1913);  there  was  then  no  reason  for  eliminating  an  example  that 
would  retain  its  didactic  value.  [Ed.] 


by  saying  that  Latin  pater  meant  the  same  thing.  Another  example : 
Latin  short  a  became  i  in  noninitial  open  syllables;  beside /acio  we 
have  conficio,  beside  amicus,  inimicus,  etc.  The  law  is  often  stated 
in  this  way:  ''The  a  of  facio  becomes  i  in  conficio  because  it  is  no 
longer  in  the  first  syllable."  That  is  not  true:  never  did  the  a 
"become"  i  in  conficio.  To  re-establish  the  truth  one  must  single  out 
two  periods  and  four  terms.  Speakers  first  said  facio — confacio; 
then,  confacio  having  been  changed  to  conficio  while  facio  remained 
unchanged,  they  said  facio — conficio: 

facio  < >  confacio        Period  A 

facio  < >  conficio         Period  B 

If  a  "change"  occurred,  it  is  between  confacio  and  conficio;  but  the 
rule,  badly  formulated,  does  not  even  mention  confacio!  Then  be- 
side the  diachronic  change  there  is  a  second  fact,  absolutely  distinct 
from  the  first  and  having  to  do  with  the  purely  synchronic  op- 
position between  facio  and  conficio.  One  is  tempted  to  say  that  it 
is  not  a  fact  but  a  result.  Nevertheless,  it  is  a  fact  in  its  own  class; 
indeed,  all  synchronic  phenomena  are  like  this.  The  true  value  of 
the  opposition  facio:  conficio  is  not  recognized  for  the  very  reason 
that  the  opposition  is  not  very  significant.  But  oppositions  like 
Gast:  Gdste  and  gebe:  gibt,  though  also  fortuitous  results  of  phonetic 
evolution,  are  nonetheless  basic  grammatical  phenomena  of  the 
synchronic  class.  The  fact  that  both  classes  are  in  other  respects 
closely  linked,  each  conditioning  the  other,  points  to  the  conclusion 
that  keeping  them  apart  is  not  worthwhile ;  in  fact,  linguistics  has 
confused  them  for  decades  without  realizing  that  such  a  method 
is  worthless. 

The  mistake  shows  up  conspicuously  in  certain  instances.  To 
explain  Greek  phuktos,  for  example,  it  might  seem  sufficient  to  say 
that  in  Greek  g  or  kh  became  k  before  voiceless  consonants,  and  to 
cite  by  way  of  explanation  such  synchronic  correspondences  as 
phugein:  phuktos,  lekhos:  lektron,  etc.  But  in  a  case  like  trikhes: 
thriksi  there  is  a  complication,  the  "passing"  of  t  to  th.  The  forms 
can  be  explained  only  historically,  by  relative  chronology.  The 
Proto-Greek  theme  Hhrikh,  followed  by  the  ending  -si,  became 
thriksi,  a  very  old  development  identical  to  the  one  that  produced 


lektron  from  the  root  lekh-.  Later  every  aspirate  followed  by  an- 
other aspirate  in  the  same  word  was  changed  into  an  occlusive,  and 
*thrikhes  became  trikhes;  naturally  thriksi  escaped  this  law. 

9.  Conclusions 

Linguistics  here  comes  to  its  second  bifurcaton.  We  had  first  to 
choose  between  language  and  speaking  (see  pp.  17  ff.);  here  we  are 
again  at  the  intersection  of  two  roads,  one  leading  to  diachrony 
and  the  other  to  synchrony. 

Once  in  possession  of  this  double  principle  of  classification,  we 
can  add  that  everything  diachronic  in  language  is  diachronic  only 
by  virtue  of  speaking.  It  is  in  speaking  that  the  germ  of  all  change 
is  found.  Each  change  is  launched  by  a  certain  number  of  indi- 
viduals before  it  is  accepted  for  general  use.  Modern  German  uses 
ich  war,  wir  waren,  whereas  until  the  sixteenth  century  the  con- 
jugation was  ich  was,  wir  waren  (cf.  English  I  was,  we  were).  How 
did  the  substitution  of  war  for  was  come  about?  Some  speakers, 
influenced  by  waren,  created  war  through  analogy;  this  was  a  fact 
of  speaking;  the  new  form,  repeated  many  times  and  accepted  by 
the  community,  became  a  fact  of  language.  But  not  all  innovations 
of  speaking  have  the  same  success,  and  so  long  as  they  remain  in- 
dividual, they  may  be  ignored,  for  we  are  studying  language ;  they 
do  not  enter  into  our  field  of  observation  until  the  community  of 
speakers  has  adopted  them. 

An  evolutionary  fact  is  always  preceded  by  a  fact,  or  rather  by 
a  multitude  of  similar  facts,  in  the  sphere  of  speaking.  This  in  no 
way  invalidates  but  rather  strengthens  the  distinction  made  above 
since  in  the  history  of  any  innovation  there  are  always  two  distinct 
moments:  (1)  when  it  sprang  up  in  individual  usage;  and  (2)  when 
it  became  a  fact  of  language,  outwardly  identical  but  adopted  by 
the  community. 

The  following  table  indicates  the  rational  form  that  linguistic 
study  should  take : 

, ,  ^^„„-  Diachrony 



One  must  recognize  that  the  ideal,  theoretical  form  of  a  science  is 
not  always  the  one  imposed  upon  it  by  the  exigencies  of  practice ; 
in  Unguistics  these  exigencies  are  more  imperious  than  anywhere 
else ;  they  account  to  some  extent  for  the  confusion  that  now  pre- 
dominates in  linguistic  research.  Even  if  the  distinctions  set  up  here 
were  accepted  once  and  for  all,  a  precise  orientation  probably  could 
not  be  imposed  on  investigations  in  the  name  of  the  stated  ideal. 

In  the  synchronic  study  of  Old  French,  for  instance,  the  hnguist 
works  with  facts  and  principles  that  have  nothing  in  common  with 
those  that  he  would  find  out  by  tracing  the  history  of  the  same 
language  from  the  thirteenth  to  the  twentieth  century;  on  the 
contrary,  he  works  with  facts  and  principles  similar  to  those  that 
would  be  revealed  in  the  description  of  an  existing  Bantu  language, 
Attic  Greek  of  400  b.c.  or  present-day  French,  for  that  matter. 
These  diverse  descriptions  would  be  based  on  similar  relations;  if 
each  idiom  is  a  closed  system,  all  idioms  embody  certain  fixed 
principles  that  the  linguist  meets  again  and  again  in  passing  from 
one  to  another,  for  he  is  staying  in  the  same  class.  Historical  study 
is  no  different.  Whether  the  linguist  examines  a  definite  period  in 
the  history  of  French  (for  example,  from  the  thirteenth  to  the 
twentieth  century)  Javanese,  or  any  other  language  whatsoever, 
everywhere  he  works  with  similar  facts  which  he  needs  only  com- 
pare in  order  to  establish  the  general  truths  of  the  diachronic  class. 
The  ideal  would  be  for  each  scholar  to  devote  himself  to  one  field 
of  investigation  or  the  other  and  deal  with  the  largest  possible 
number  of  facts  in  this  class;  but  it  is  very  difficult  to  command 
scientifically  such  different  languages.  Against  this,  each  language 
in  practice  forms  a  unit  of  study,  and  we  are  induced  by  force  of 
circumstances  to  consider  it  alternately  from  the  historical  and 
static  viewpoints.  Above  all  else,  we  must  never  forget  that  this 
unit  is  superficial  in  theory,  whereas  the  diversity  of  idioms  hides 
a  profound  unity.  Whichever  way  we  look  in  studying  a  language, 
we  must  put  each  fact  in  its  own  class  and  not  confuse  the  two 

The  two  parts  of  linguistics  respectively,  as  defined,  will  be  the 
object  of  our  study. 

Synchronic  linguistics  will  be  concerned  with  the  logical  and 


psychological  relations  that  bind  together  coexisting  terms  and 
form  a  system  in  the  collective  mind  of  speakers. 

Diachronic  linguistics,  on  the  contrary,  will  study  relations  that 
bind  together  successive  terms  not  perceived  by  the  collective  mind 
but  substituted  for  each  other  without  forming  a  system. 


Synchronic  Linguistics 

Chapter  I 

The  aim  of  general  synchronic  linguistics  is  to  set  up  the  funda- 
mental principles  of  any  idiosynchronic  system,  the  constituents 
of  any  language-state.  Many  of  the  items  already  explained  in  Part 
One  belong  rather  to  synchrony ;  for  instance,  the  general  properties 
of  the  sign  are  an  integral  part  of  synchrony  although  they  were 
used  to  prove  the  necessity  of  separating  the  two  linguistics. 

To  synchrony  belongs  everything  called  "general  grammar," 
for  it  is  only  through  language-states  that  the  different  relations 
which  are  the  province  of  grammar  are  established.  In  the  following 
chapters  we  shall  consider  only  the  basic  principles  necessary  for 
approaching  the  more  special  problems  of  static  linguistics  or 
explaining  in  detail  a  language-state. 

The  study  of  static  linguistics  is  generally  much  more  difficult 
than  the  study  of  historical  linguistics.  Evolutionary  facts  are  more 
concrete  and  striking ;  their  observable  relations  tie  together  succes- 
sive terms  that  are  easily  grasped ;  it  is  easy,  often  even  amusing,  to 
follow  a  series  of  changes.  But  the  linguistics  that  penetrates 
values  and  coexisting  relations  presents  much  greater  difficulties. 

In  practice  a  language-state  is  not  a  point  but  rather  a  certain 
span  of  time  during  which  the  sum  of  the  modifications  that  have 
supervened  is  minimal.  The  span  may  cover  ten  years,  a  gener- 
ation, a  century,  or  even  more.  It  is  possible  for  a  language  to 
change  hardly  at  all  over  a  long  span  and  then  to  undergo  radical 
transformations  within  a  few  years.  Of  two  languages  that  exist 
side  by  side  during  a  given  period,  one  may  evolve  drastically  and 
the  other  practically  not  at  all;  study  would  have  to  be  diachronic 
in  the  former  instance,  synchronic  in  the  latter.  An  absolute  state 
is  defined  by  the  absence  of  changes,  and  since  language  changes 



somewhat  in  spite  of  everything,  studying  a  language-state  means 
in  practice  disregarding  changes  of  little  importance,  just  as 
mathematicians  disregard  infinitesimal  quantities  in  certain  cal- 
culations, such  as  logarithms. 

Political  history  makes  a  distinction  between  era,  a  point  in  time, 
and  period,  which  embraces  a  certain  duration.  Still,  the  historian 
speaks  of  the  Antoninian  Era,  the  Era  of  the  Crusades,  etc.  when 
he  considers  a  set  of  characteristics  which  remained  constant  dur- 
ing those  times.  One  might  also  say  that  static  linguistics  deals  with 
eras.  But  state  is  preferable.  The  beginning  and  the  end  of  an  era 
are  generally  characterized  by  some  rather  brusque  revolution  that 
tends  to  modify  the  existing  state  of  affairs.  The  word  state  avoids 
giving  the  impression  that  anything  similar  occurs  in  language. 
Besides,  precisely  because  it  is  borrowed  from  history,  the  term  era 
makes  one  think  less  of  language  itself  than  of  the  circumstances 
that  surround  it  and  condition  it;  in  short,  it  suggests  rather  the 
the  idea  of  what  we  called  external  linguistics  (see  p.  20) . 

Besides,  delimitation  in  time  is  not  the  only  difficulty  that  we 
encounter  in  defining  a  language-state:  space  presents  the  same 
problem.  In  short,  a  concept  of  a  language-state  can  be  only  ap- 
proximate. In  static  linguistics,  as  in  most  sciences,  no  course  of 
reasoning  is  possible  without  the  usual  simplification  of  data. 

Chapter  II 

1 .  Definition:  Entity  and  Unit 
The  signs  that  make  up  language  are  not  abstractions  but  real 

objects  (see  p.  15);  signs  and  their  relations  are  what  linguistics 

studies;  they  are  the  concrete  entities  of  our  science. 

Let  us  first  recall  two  principles  that  dominate  the  whole  issue : 
1)  The  linguistic  entity  exists  only  through  the  associating  of  the 

signifier  with  the  signified  (see  p.  66  ff.).  Whenever  only  one  ele- 


ment  is  retained,  the  entity  vanishes;  instead  of  a  concrete  object 
we  are  faced  with  a  mere  abstraction.  We  constantly  risk  grasping 
only  a  part  of  the  entity  and  thinking  that  we  are  embracing  it  in 
its  totality;  this  would  happen,  for  example,  if  we  divided  the 
spoken  chain  into  syllables,  for  the  syllable  has  no  value  except  in 
phonology.  A  succession  of  sounds  is  linguistic  only  if  it  supports 
an  idea.  Considered  independently,  it  is  material  for  a  physiologi- 
cal study,  and  nothing  more  than  that. 

The  same  is  true  of  the  signified  as  soon  as  it  is  separated  from 
its  signifier.  Considered  independently,  concepts  like  "house," 
"white,"  "see,"  etc.  belong  to  psychology.  They  become  linguistic 
entities  only  when  associated  with  sound-images;  in  language,  a 
concept  is  a  quality  of  its  phonic  substance  just  as  a  particular 
slice  of  sound  is  a  quality  of  the  concept. 

The  two-sided  linguistic  unit  has  often  been  compared  with  the 
human  person,  made  up  of  the  body  and  the  soul.  The  comparison 
is  hardly  satisfactory.  A  better  choice  would  be  a  chemical  com- 
pound like  water,  a  combination  of  hydrogen  and  oxygen;  taken 
separately,  neither  element  has  any  of  the  properties  of  water. 

2)  The  Hnguistic  entity  is  not  accurately  defined  until  it  is 
delimited,  i.e.  separated  from  everything  that  surrounds  it  on  the 
phonic  chain.  These  delimited  entities  or  units  stand  in  opposition 
to  each  other  in  the  mechanism  of  language. 

One  is  at  first  tempted  to  hken  linguistic  signs  to  visual  signs, 
which  can  exist  in  space  without  becoming  confused,  and  to  assume 
that  separation  of  the  significant  elements  can  be  accomplished  in 
the  same  way,  without  recourse  to  any  mental  process.  The  word 
"form,"  which  is  often  used  to  indicate  them  (cf.  the  expression 
"verbal  form,"  "noun  form")  gives  support  to  the  mistake.  But 
we  know  that  the  main  characteristic  of  the  sound-chain  is  that  it 
is  linear  (see  p.  70).  Considered  by  itself,  it  is  only  a  line,  a  con- 
tinuous ribbon  along  which  the  ear  perceives  no  self-sufficient  and 
clear-cut  division;  to  divide  the  chain,  we  must  call  in  meanings. 
When  we  hear  an  unfamiliar  language,  we  are  at  a  loss  to  say  how 
the  succession  of  sounds  should  be  analyzed,  for  analysis  is  impos- 
sible if  only  the  phonic  side  of  the  linguistic  phenomenon  is  con- 
sidered. But  when  we  know  the  meaning  and  function  that  must 



be  attributed  to  each  part  of  the  chain,  we  see  the  parts  detach 
themselves  from  each  other  and  the  shapeless  ribbon  break  into 
segments.  Yet  there  is  nothing  material  in  the  analysis. 

To  summarize:  language  does  not  offer  itself  as  a  set  of  pre- 
delimited  signs  that  need  only  be  studied  according  to  their  mean- 
ing and  arrangement ;  it  is  a  confused  mass,  and  only  attentiveness 
and  familiarization  will  reveal  its  particular  elements.  The  unit  has 
no  special  phonic  character,  and  the  only  definition  that  we  can 
give  it  is  this:  it  is  a  slice  of  sound  which  to  the  exclusion  of  everything 
that  precedes  and  follows  it  in  the  spoken  chain  is  the  signifier  of  a 
certain  concept. 

2.  Method  of  Delimitation 

One  who  knows  a  language  singles  out  its  units  by  a  very  simple 
method — in  theory,  at  any  rate.  His  method  consists  of  using 
speaking  as  the  source  material  of  language  and  picturing  it  as  two 
parallel  chains,  one  of  concepts  {A)  and  the  other  of  sound-images 

In  an  accurate  delimitation,  the  division  along  the  chain  of 
sound-images  (a,  h,  c)  will  correspond  to  the  division  along  the 
chain  of  concepts  (a',  h',  c') : 



Take  French  sizlapra.  Can  we  cut  the  chain  after  I  and  make  sizl 
a  unit?  No,  we  need  only  consider  the  concepts  to  see  that  the 
division  is  wrong.  Neither  is  the  syllabic  division  siz-la-pra  to  be 
taken  for  granted  as  having  linguistic  value.  The  only  possible 
divisions  are  these:  (1)  si-z-la-pra  (si  je  la  prends  'if  I  take  it')  and 
(2)  si-z-l-apra  (si  je  Vapprends  'if  I  learn  it'),  and  they  are  deter- 
mined by  the  meaning  that  is  attached  to  the  words. ^ 

To  verify  the  result  of  the  procedure  and  be  assured  that  we  are 
really  deahng  with  a  unit,  we  must  be  able  in  comparing  a  series  of 

*  Cf.  the  sounds  [jurmam]  in  English:  "your  mine"  or  "you're  mine."  [Tr.] 


sentences  in  which  the  same  unit  occurs  to  separate  the  unit  from 
the  rest  of  the  context  and  find  in  each  instance  that  meaning  jus- 
tifies the  delimitation.  Take  the  two  French  phrases  lafprsdiiva 
(la,  force  du  vent  'the  force  of  the  wind'),  and  abudfgrs  (a  bout  de 
force  'exhausted';  literally:  'at  the  end  of  one's  force').  In  each 
phrase  the  same  concept  coincides  with  the  same  phonic  sHce,  fgrs; 
thus  it  is  certainly  a  linguistic  unit.  But  in  ilmdfgrsaparle  (il  me 
force  a  parler  'he  forces  me  to  talk')  fors  has  an  entirely  different 
meaning:  it  is  therefore  another  unit. 

3.  Practical  Difficulties  of  Delimitation 

The  method  outlined  above  is  very  simple  in  theory,  but  is  it 
easy  to  apply?  We  are  tempted  to  think  so  if  we  start  from  the 
notion  that  the  units  to  be  isolated  are  words.  For  what  is  a  sen- 
tence except  a  combination  of  words?  And  what  can  be  grasped 
more  readily  than  words?  Going  back  to  the  example  given  above, 
we  may  say  that  the  analysis  of  the  spoken  chain  sizlaprd  resulted 
in  the  delimiting  of  four  units,  and  that  the  units  are  words :  si-je-l- 
apprends.  But  we  are  immediately  put  on  the  defensive  on  noting 
that  there  has  been  much  disagreement  about  the  nature  of  the 
word,  and  a  little  reflection  shows  that  the  usual  meaning  of  the 
term  is  incompatible  with  the  notion  of  concrete  unit. 

To  be  convinced,  we  need  only  think  of  French  cheval  'horse'  and 
its  plural  from  chevaux.  People  readily  say  that  they  are  two  forms 
of  the  same  word ;  but  considered  as  wholes,  they  are  certainly  two 
distinct  things  with  respect  to  both  meaning  and  sound.  In 
mwa  (mois,  as  in  le  mois  de  Septembre  'the  month  of  September') 
and  mwaz  (mois,  in  un  mois  apres  *a  month  later')  there  are  also 
two  forms  of  the  same  word,  and  there  is  no  question  of  a  concrete 
unit.  The  meaning  is  the  same,  but  the  slices  of  sound  are  dif- 
ferent. As  soon  as  we  try  to  liken  concrete  units  to  words,  we 
face  a  dilemma:  we  must  either  ignore  the  relation — which  is  none- 
theless evident — that  binds  cheval  and  chevaux,  the  two  sounds  of 
mwa  and  mwaz,  etc.  and  say  that  they  are  different  words,  or  in- 
stead of  concrete  units  be  satisfied  with  the  abstraction  that  links 
the  different  forms  of  the  same  word.  The  concrete  unit  must  be 
sought,  not  in  the  word,  but  elsewhere.  Besides,  many  words  are 


complex  units,  and  we  can  easily  single  out  their  subunits  (suffixes, 
prefixes,  radicals).  Derivatives  like  pain-ful  and  delight-ful  can  be 
divided  into  distinct  parts,  each  having  an  obvious  meaning  and 
function.  Conversely,  some  units  are  larger  than  words:  compounds 
(French  porte-plume  'penholder'),  locutions  (s'il  vous  plait  'please'), 
inflected  forms  {il  a  ete  'he  has  been'),  etc.  But  these  units  resist  de- 
limitation as  strongly  as  do  words  proper,  making  it  extremely 
difficult  to  disentangle  the  interplay  of  units  that  are  found  in  a 
sound-chain  and  to  specify  the  concrete  elements  on  which  a 
language  functions. 

Doubtless  speakers  are  unaware  of  the  practical  difficulties  of 
delimiting  units.  Anything  that  is  of  even  the  slightest  significance 
seems  like  a  concrete  element  to  them  and  they  never  fail  to  single 
it  out  in  discourse.  But  it  is  one  thing  to  feel  the  quick,  delicate 
interplay  of  units  and  quite  another  to  account  for  them  through 
methodical  analysis. 

A  rather  widely  held  theory  makes  sentences  the  concrete  units 
of  language:  we  speak  only  in  sentences  and  subsequently  single 
out  the  words.  But  to  what  extent  does  the  sentence  belong  to 
language  (see  p.  124)?  If  it  belongs  to  speaking,  the  sentence  can- 
not pass  for  the  Unguistic  unit.  But  let  us  suppose  that  this  diffi- 
culty is  set  aside.  If  we  picture  to  ourselves  in  their  totality  the 
sentences  that  could  be  uttered,  their  most  striking  characteristic  is 
that  in  no  way  do  they  resemble  each  other.  We  are  at  first  tempted 
to  liken  the  immense  diversity  of  sentences  to  the  equal  diversity  of 
the  individuals  that  make  up  a  zoological  species.  But  this  is  an 
illusion :  the  characteristics  that  animals  of  the  same  species  have 
in  common  are  much  more  significant  than  the  differences  that 
separate  them.  In  sentences,  on  the  contrary,  diversity  is  domi- 
nant, and  when  we  look  for  the  link  that  bridges  their  diversity, 
again  we  find,  without  having  looked  for  it,  the  word  with  its  gram- 
matical characteristics  and  thus  fall  back  into  the  same  difficulties 
as  before. 

4.  Conclusion 

In  most  sciences  the  question  of  units  never  even  arises :  the  units 
are  delimited  from  the  outset.  In  zoology,  the  animal  immediately 
presents  itself.  Astronomy  works  with  units  that  are  separated  in 


space,  the  stars.  The  chemist  can  study  the  nature  and  composition 
of  potassium  bichromate  without  doubting  for  an  instant  that  this 
is  a  well-defined  object. 

When  a  science  has  no  concrete  units  that  are  immediately  recog- 
nizable, it  is  because  they  are  not  necessary.  In  history,  for  ex- 
ample, is  the  unit  the  individual,  the  era,  or  the  nation?  We  do  not 
know.  But  what  does  it  matter?  We  can  study  history  without 
knowing  the  answ^er. 

But  just  as  the  game  of  chess  is  entirely  in  the  combination  of 
the  different  chesspieces,  language  is  characterized  as  a  system 
based  entirely  on  the  opposition  of  its  concrete  units.  We  can 
neither  dispense  with  becoming  acquainted  with  them  nor  take  a 
single  step  without  coming  back  to  them;  and  still,  delimiting  them 
is  such  a  dehcate  problem  that  we  may  wonder  at  first  whether 
they  really  exist. 

Language  then  has  the  strange,  striking  characteristic  of  not 
having  entities  that  are  perceptible  at  the  outset  and  yet  of  not 
permitting  us  to  doubt  that  they  exist  and  that  their  functioning 
constitutes  it.  Doubtless  we  have  here  a  trait  that  distinguishes 
language  from  all  other  semiological  institutions. 

Chapter  III 


The  statement  just  made  brings  us  squarely  up  against  a  problem 
that  is  all  the  more  important  because  any  basic  notion  in  static 
linguistics  depends  directly  on  our  conception  of  the  unit  and  even 
blends  w^ith  it.  This  is  what  I  should  like  successively  to  dem- 
onstrate with  respect  to  the  notions  of  synchronic  identity,  reality, 
and  value. 

A.  What  is  a  synchronic  identity f  Here  it  is  not  a  question  of  the 
identity  that  links  the  French  negation  pas  'not'  to  Latin  passum, 
a  diachronic  identity  that  will  be  dealt  with  elsewhere  (see  p.  181), 
but  rather  of  the  equally  interesting  identity  by  virtue  of  which  we 


state  that  two  sentences  like  je  ne  sais  jpas  'I  don't  know'  and  ne 
dttes  yas  cela  ^don't  say  that'  contain  the  same  element.  An  idle 
question,  one  might  say;  there  is  identity  because  the  same  slice  of 
sound  carries  the  same  meaning  in  the  two  sentences.  But  that 
explanation  is  unsatisfactory,  for  if  the  correspondence  of  slices  of 
sound  and  concepts  is  proof  of  identity  (see  above,  p.  105,  la  force 
du  vent :  a  bout  de  force) ,  the  reverse  is  not  true.  There  can  be 
identity  without  this  correspondence.  When  Gentlemen!  is  repeated 
several  times  during  a  lecture,  the  listener  has  the  feeling  that  the 
same  expression  is  being  used  each  time,  and  yet  variations  in 
utterance  and  intonation  make  for  appreciable  phonic  differences 
in  diverse  contexts — differences  just  as  appreciable  as  those  that 
elsewhere  separate  different  words  (cf.  French  pomme  'apple'  and 
paume  'palm,'  goutte  'drop'  and  je  goute  'I  taste,'  fuir  'flee,'  and 
fouir  'stuff,'  etc.);2  besides,  the  feeling  of  identity  persists  even 
though  there  is  no  absolute  identity  between  one  Gentlemen!  and 
the  next  from  a  semantic  viewpoint  either.  In  the  same  vein,  a 
word  can  express  quite  different  ideas  without  compromising  its 
identity  (cf.  French  adopter  une  mode  'adopt  a  fashion'  and  adopter 
un  enfant  'adopt  a  child,'  la  fleur  du  pommier  'the  flower  of  the 
apple  tree'  and  la,  fleur  de  la  noblesse  'the  flower  of  nobility,'  etc.). 
The  Unguistic  mechanism  is  geared  to  differences  and  identities, 
the  former  being  only  the  counterpart  of  the  latter.  Everjrwhere 
then,  the  problem  of  identities  appears;  moreover,  it  blends  par- 
tially with  the  problem  of  entities  and  units  and  is  only  a  compH- 
cation — illuminating  at  some  points — of  the  larger  problem.  This 
characteristic  stands  out  if  we  draw  some  comparisons  with  facts 
taken  from  outside  speech.  For  instance,  we  speak  of  the  identity  of 
two  "8:25  p.m.  Geneva-to-Paris"  trains  that  leave  at  twenty-four 
hour  intervals.  We  feel  that  it  is  the  same  train  each  day,  yet  every- 
thing— the  locomotive,  coaches,  personnel — is  probably  different. 
Or  if  a  street  is  demolished,  then  rebuilt,  we  say  that  it  is  the  same 
street  even  though  in  a  material  sense,  perhaps  nothing  of  the  old 
one  remains.  Why  can  a  street  be  completely  rebuilt  and  still  be 
the  same?  Because  it  does  not  constitute  a  purely  material  entity ; 
it  is  based  on  certain  conditions  that  are  distinct  from  the  materials 

^  Cf.  English  bought:  boat,  naught:  note,  far:  for:  four  (for  many  speakers). 


that  fit  the  conditions,  e.g.  its  location  with  respect  to  other  streets. 
Similarly,  what  makes  the  express  is  its  hour  of  departure,  its 
route,  and  in  general  every  circumstance  that  sets  it  apart  from 
other  trains.  Whenever  the  same  conditions  are  fulfilled,  the  same 
entities  are  obtained.  Still,  the  entities  are  not  abstract  since  we 
cannot  conceive  of  a  street  or  train  outside  its  material  reahzation. 

Let  us  contrast  the  preceding  examples  with  the  completely 
different  case  of  a  suit  which  has  been  stolen  from  me  and  which  I 
find  in  the  window  of  a  second-hand  store.  Here  we  have  a  material 
entity  that  consists  solely  of  the  inert  substance — the  cloth,  its 
lining,  its  trimmings,  etc.  Another  suit  would  not  be  mine  regard- 
less of  its  similarity  to  it.  But  linguistic  identity  is  not  that  of  the 
garment;  it  is  that  of  the  train  and  the  street.  Each  time  I  say  the 
word  Gentlemen!  I  renew  its  substance;  each  utterance  is  a  new 
phonic  act  and  a  new  psychological  act.  The  bond  between  the  two 
uses  of  the  same  word  depends  neither  on  material  identity  nor  on 
sameness  in  meaning  but  on  elements  which  must  be  sought  after 
and  which  will  point  up  the  true  nature  of  linguistic  units. 

B.  What  is  a  sjmchronic  reality?  To  what  concrete  or  abstract 
elements  of  language  can  the  name  be  applied? 

Take  as  an  example  the  distinction  between  the  parts  of  speech. 
What  supports  the  classing  of  words  as  substantives,  adjectives, 
etc.?  Is  it  done  in  the  name  of  a  purely  logical,  extra-linguistic 
principle  that  is  applied  to  grammar  from  without  like  the  degrees 
of  longitude  and  latitude  on  the  globe?  Or  does  it  correspond  to 
something  that  has  its  place  in  the  system  of  language  and  is  con- 
ditioned by  it?  In  a  word,  is  it  a  synchronic  reality?  The  second 
supposition  seems  probable,  but  the  first  could  also  be  defended. 
In  the  French  sentence  ces  gants  sont  hon  marche  'these  gloves  are 
cheap,'  is  hon  marche  an  adjective?  It  is  apparently  an  adjective 
from  a  logical  viewpoint  but  not  from  the  viewpoint  of  grammar, 
for  hon  marche  fails  to  behave  as  an  adjective  (it  is  invariable,  it 
never  precedes  its  noun,  etc.);  in  addition,  it  is  composed  of  two 
words.  Now  the  distinction  between  parts  of  speech  is  exactly  what 
should  serve  to  classify  the  words  of  language.  How  can  a  group  of 
words  be  attributed  to  one  of  the  "parts"?  But  to  say  that  hon 
'good'  is  an  adjective  and  marche  'market'  a  substantive  explains 
nothing.  We  are  then  dealing  with  a  defective  or  incomplete  clas- 


sification ;  the  division  of  words  into  substantives,  verbs,  adjectives, 
etc.  is  not  an  undeniable  linguistic  reality.' 

Linguistics  accordingly  works  continuously  with  concepts  forged 
by  grammarians  without  knowing  whether  or  not  the  concepts 
actually  correspond  to  the  constituents  of  the  system  of  language. 
But  how  can  we  find  out?  And  if  they  are  phantoms,  what  realities 
can  we  place  in  opposition  to  them? 

To  be  rid  of  illusions  we  must  first  be  convinced  that  the  con^ 
Crete  entities  of  language  are  not  directly  accessible.  If  we  try  to^ 
grasp  them,  we  come  into  contact  with  the  true  facts. -Starting 
Mi;om  there,  we  can  set  up  all  the  clasgifi^atioiis  that  linguistics 

nee^S'fui'  uii  anglilg  all  fheTaCts^'Tts  disposaly(5n  the  ofheiTiand^"'^ 
^.tor  base  the  classifications  on  anything  except  concrete  entities — to 
[  say,  for  example,  that  the  parts  of  speech  are  the  constituents  of 
\  language  simply  because  they  correspond  to  categories  of  logic — is 
\  to  forget  that  there  are  no  linguistic  facts  apart  from  the  phonic 
Nmbstance  cut  into  significant  elements. 

X>.  Finally,  not  every  idea  touched  upon  in  this  chapter  differs 
basically  from  what  we  have  elsewhere  called  values.  A  new  com- 
parison with  the  set  of  chessmen  will  bring  out  this  point  (see 
pp.  88  ff.).  Take  a  knight,  for  instance.  By  itself  is  it  an  element  in 
the  game?  Certainly  not,  for  by  its  material  make-up — outside  its 
square  and  the  other  conditions  of  the  game — it  means  nothing  to 
the  player;  it  becomes  a  real,  concrete  element  only  when  endowed 
with  value  and  wedded  to  it.  Suppose  that  the  piece  happens  to  be 
destroyed  or  lost  during  a  game.  Can  it  be  replaced  by  an  equiva- 
lent piece?  Certainly.  Not  only  another  knight  but  even  a  figure 
shorn  of  any  resemblance  to  a  knight  can  be  declared  identical 
provided  the  same  value  is  attributed  to  it.  We  see  then  that  in 
semiological  systems  like  language,  where  elements  hold  each  other 
in  equilibrium  in  accordance  with  fixed  rules,  the  notion  of  identity 
blends  with  that  of  value  and  vice  versa. 

In  a  word,  that  is  why  the  notion  of  value  envelopes  the  notions 
of  unit,  concrete  entity,  and  reality.  But  if  there  is  no  fundamental 

*  Form,  function,  and  meaning  combine  to  make  the  classing  of  the  parts  of 
speech  even  more  difficult  in  English  than  in  French.  Cf.  ten-foot:  ten  feet  in 
a  ten-foot  pole:  the  pole  is  ten  feet  long.  [Tr.] 


difference  between  these  diverse  notions,  it  follows  that  the  prob- 
lem can  be  stated  successively  in  several  ways.  Whether  we  try  to 
define  the  unit,  reality,  concrete  entity,  or  value,  we  always  come 
back  to  the  central  question  that  dominates  all  of  static  linguistics. 

It  would  be  interesting  from  a  practical  viewpoint  to  begin  with 
units,  to  determine  what  they  are  and  to  account  for  their  diversity 
by  classifying  them.  It  would  be  necessary  to  search  for  the  reason 
for  dividing  language  into  words — for  in  spite  of  the  difficulty  of 
defining  it,  the  word  is  a  unit  that  strikes  the  mind,  something 
central  in  the  mechanism  of  language — but  that  is  a  subject  which 
by  itself  would  fill  a  volume.  Next  we  would  have  to  classify  the 
subunits,  then  the  larger  units,  etc.  By  determining  in  this  way 
the  elements  that  it  manipulates,  synchronic  linguistics  would 
completely  fulfill  its  task,  for  it  would  relate  all  synchronic  phe- 
nomena to  their  fundamental  principle.  It  cannot  be  said  that  this 
basic  problem  has  ever  been  faced  squarely  or  that  its  scope  and 
difficulty  have  been  understood ;  in  the  matter  of  language,  people 
have  always  been  satisfied  with  ill-defined  units. 

Still,  in  spite  of  their  capital  importance,  it  is  better  to  approach 
the  problem  of  units  through  the  study  of  value,  for  in  my  opinion 
value  is  of  prime  importance. 

Chapter  IV 


1.  Language  as  Organized  Thought  Coupled  with  Sound 

To  prove  that  language  is  only  a  system  of  pure  values,  it  is 
enough  to  consider  the  two  elements  involved  in  its  functioning: 
ideas  and  sounds. 

Psychologically  our  thought — apart  from  its  expression  in  words 
— is  only  a  shapeless  and  indistinct  mass.  Philosophers  and  lin- 
guists have  always  agreed  in  recognizing  that  without  the  help  of 
signs  we  would  be  unable  to  make  a  clear-cut,  consistent  distinction 



between  two  ideas.  Without  language,  thought  is  a  vague,  un- 
charted nebula.  There  are  no  pre-existing  ideas,  and  nothing  is 
distinct  before  the  appearance  of  language. 

Against  the  floating  realm  of  thought,  would  sounds  by  them- 
selves yield  predelimited  entities?  No  more  so  than  ideas.  Phonic 
substance  is  neither  more  fixed  nor  more  rigid  than  thought;  it  is 
not  a  mold  into  which  thought  must  of  necessity  fit  but  a  plastic 
substance  divided  in  turn  into  distinct  parts  to  furnish  the  signifiers 
needed  by  thought.  The  linguistic  fact  can  therefore  be  pictured 
in  its  totality — i.e.  language — as  a  series  of  contiguous  subdivisions 
marked  off  on  both  the  indefinite  plane  of  jumbled  ideas  (A)  and 
the  equally  vague  plane  of  sounds  (B).  The  following  diagram 
gives  a  rough  idea  of  it : 

The  characteristic  role  of  language  with  respect  to  thought  is  not 
to  create  a  material  phonic  means  for  expressing  ideas  but  to  serve 
as  a  link  between  thought  and  sound,  under  conditions  that 
of  necessity  bring  about  the  reciprocal  delimitations  of  units. 
Thought,  chaotic  by  nature,  has  to  become  ordered  in  the  process 
of  its  decomposition.  Neither  are  thoughts  given  material  form 
nor  are  sounds  transformed  into  mental  entities;  the  somewhat 
mysterious  fact  is  rather  that  "thought-sound"  implies  division, 
and  that  language  works  out  its  units  while  taking  shape  between 
two  shapeless  masses.  Visualize  the  air  in  contact  with  a  sheet  of 
water;  if  the  atmospheric  pressure  changes,  the  surface  of  the 
water  will  be  broken  up  into  a  series  of  divisions,  waves;  the  waves 
resemble  the  union  or  coupUng  of  thought  with  phonic  substance. 

Language  might  be  called  the  domain  of  articulations,  using  the 


word  as  it  was  defined  earlier  (see  p.  10).  Each  linguistic  term  is  a 
member,  an  articulus  in  which  an  idea  is  fixed  in  a  sound  and  a 
sound  becomes  the  sign  of  an  idea. 

Language  can  also  be  compared  with  a  sheet  of  paper:  thought 
is  the  front  and  the  sound  the  back;  one  cannot  cut  the  front  with- 
out cutting  the  back  at  the  same  time;  likewise  in  language,  one 
can  neither  divide  sound  from  thought  nor  thought  from  sound; 
the  division  could  be  accomplished  only  abstractedly,  and  the 
result  would  be  either  pure  psychology  or  pure  phonology. 

Linguistics  then  works  in  the  borderland  where  the  elements  of 
sound  and  thought  combine ;  their  combination  produces  a  form,  not 
a  substance. 

These  views  give  a  better  understanding  of  what  was  said  before 
(see  pp.  67  ff.)  about  the  arbitrariness  of  signs.  Not  only  are  the  two 
domains  that  are  linked  by  the  linguistic  fact  shapeless  and  con- 
fused, but  the  choice  of  a  given  slice  of  sound  to  name  a  given  idea 
is  completely  arbitrary.  If  this  were  not  true,  the  notion  of  value 
would  be  compromised,  for  it  would  include  an  externally  imposed 
element.  But  actually  values  remain  entirely  relative,  and  that  is 
why  the  bond  between  the  sound  and  the  idea  is  radicallv 

arbitrary.  "^      ~         ^ 

/The  arbitrary  nature  of  the  sign  explains  in  turn  why  the  social 
'fact  alone  can  create  a  Hnguistic  system.  The  community  is  neces- 
sary if  values  that  owe  their  existence  solely  to  usage  and  general 
acceptance  are  to  be  set  up ;  by  himself  the  individual  is  incapable 

In  addition,  the  idea  of  value,  as  defined,  shows  that  to  consider 
a  term  as  simply  the  union  of  a  certain  sound  with  a  certain  concept 
is  grossly  misleading.  To  define  it  in  this  way  would  isolate  the 
term  from  its  system;  it  would  mean  assuming  that  one  can  start 
from  the  terms  and  construct  the  system  by  adding  them  together 
when,  on  the  contrary,  it  is  from  the  interdependent  whole  that 
one  must  start  and  through  analysis  obtain  its  elements. 

To  develop  this  thesis,  we  shall  study  value  successively  from 
the  viewpoint  of  the  signified  or  concept  (Section  2),  the  signifier 
(Section  3),  and  the  complete  sign  (Section  4). 

Being  unable  to  seize  the  concrete  entities  or  units  of  language 
directly,  we  shall  work  with  words.  While  the  word  does  not  con- 



form  exactly  to  the  definition  of  the  Hnguistic  unit  (see  p.  105), 
it  at  least  bears  a  rough  resemblance  to  the  unit  and  has  the  ad- 
vantage of  being  concrete;  consequently,  we  shall  use  words  as 
specimens  equivalent  to  real  terms  in  a  synchronic  system,  and  the 
principles  that  we  evolve  with  respect  to  words  will  be  vaUd  for 
entities  in  general. 

2.  Linguistic  Value  from  a  Conceptual  Viewpoint 

When  we  speak  of  the  value  of  a  word,  we  generally  think  first  of 
its  property  of  standing  for  an  idea,  and  this  is  in  fact  one  side  of 
linguistic  value.  But  if  this  is  true,  how  does  value  differ  from 
signification?  Might  the  two  words  be  synonyms?  I  think  not, 
although  it  is  easy  to  confuse  them,  since  the  confusion  results  not 
so  much  from  their  similarity  as  from  the  subtlety  of  the  distinction 
that  they  mark. 

From  a  conceptual  viewpoint,  value  is  doubtless  one  element  in 
signification,  and  it  is  difficult  to  see  how  signification  can  be  de- 
pendent upon  value  and  still  be  distinct  from  it.  But  we  must  clear 
up  the  issue  or  risk  reducing  language  to  a  simple  naming-process 
(see  p.  65). 

Let  us  first  take  signification  as  it  is  generally  understood  and  as 
it  was  pictured  on  page  67.  As  the  arrows  in  the  drawing  show,  it  is 
only  the  counterpart  of  the  sound-image.  Everything  that  occurs 
concerns  only  the  sound-image  and  the  concept  when  we  look  upon 
the  word  as  independent  and  self-contained. 

But  here  is  the  paradox :  on  the  one  hand  the  concept  seems  to  be 
the  counterpart  of  the  sound-image,  and  on  the  other  hand  the  sign 
itself  is  in  turn  the  counterpart  of  the  other  signs  of  language. 

Language  is  a  system  of  interdependent  terms  in  which  the 
value  of  each  term  results  solely  from  the  simultaneous  presence 
of  the  others,  as  in  the  diagram: 


How,  then,  can  value  be  confused  with  signification,  i.e.  the  coun- 
terpart of  the  sound-image?  It  seems  impossible  to  liken  the  rela- 
tions represented  here  by  horizontal  arrows  to  those  represented 
above  (p.  114)  by  vertical  arrows.  Putting  it  another  way — and 
again  taking  up  the  example  of  the  sheet  of  paper  that  is  cut  in  two 
(see  p.  1 13) — it  is  clear  that  the  observable  relation  between  the  dif- 
ferent pieces  A,  B,  C,  D,  etc.  is  distinct  from  the  relation  between 
the  front  and  back  of  the  same  piece  as  in  A/A',  B/B',  etc. 

To  resolve  the  issue,  let  us  observe  from  the  outset  that  even 
outside  language  all  values  are  apparently  governed  by  the  same 
paradoxical  principle.  They  are  always  composed: 

(1)  of  a  dissimilar  thing  that  can  be  exchanged  for  the  thing  of 
which  the  value  is  to  be  determined ;  and 

(2)  of  similar  things  that  can  be  compared  with  the  thing  of 
which  the  value  is  to  be  determined. 

Both  factors  are  necessary  for  the  existence  of  a  value.  To  de- 
termine what  a  five-franc  piece  is  worth  one  must  therefore  know : 
(1)  that  it  can  be  exchanged  for  a  fixed  quantity  of  a  different  thing, 
e.g.  bread;  and  (2)  that  it  can  be  compared  with  a  similar  value  of 
the  same  system,  e.g.  a  one-franc  piece,  or  with  coins  of  another 
system  (a  dollar,  etc.).  In  the  same  way  a  word  can  be  exchanged 
for  something  dissimilar,  an  idea ;  besides,  it  can  be  compared  with 
something  of  the  same  nature,  another  word.  Its  value  is  therefore 
not  fixed  so  long  as  one  simply  states  that  it  can  be  "exchanged" 
for  a  given  concept,  i.e.  that  it  has  this  or  that  signification:  one 
must  also  compare  it  with  similar  values,  with  other  words  that 
stand  in  opposition  to  it.  Its  content  is  really  fixed  only  by  the 
concurrence  of  everything  that  exists  outside  it.  Being  part  of  a 
system,  it  is  endowed  not  only  with  a  signification  but  also  and 
especially  with  a  value,  and  this  is  something  quite  different. 

A  few  examples  will  show  clearly  that  this  is  true.  Modern 
French  mouton  can  have  the  same  signification  as  English  sheep 
but  not  the  same  value,  and  this  for  several  reasons,  particularly 
because  in  speaking  of  a  piece  of  meat  ready  to  be  served  on  the 


table,  English  uses  mutton  and  not  sheep.  The  difference  in  value 
between  sheep  and  mouton  is  due  to  the  fact  that  sheep  has  beside 
it  a  second  term  while  the  French  word  does  not. 

Within  the  same  language,  all  words  used  to  express  related 
ideas  limit  each  other  reciprocally;  synonyms  like  French  redouter 
'dread,'  craindre  'fear,'  and  avoir  peur  'be  afraid'  have  value  only 
through  their  opposition:  if  redouter  did  not  exist,  all  its  content 
would  go  to  its  competitors.  Conversely,  some  words  are  enriched 
through  contact  with  others:  e.g.  the  new  element  introduced  in 
decrepit  (un  vieillard  decripit,  see  p.  83)  results  from  the  co- 
existence of  decrepi  (un  mur  decrepi).  The  value  of  just  any  term 
is  accordingly  determined  by  its  environment;  it  is  impossible  to 
fix  even  the  value  of  the  word  signifying  "sun"  without  first  con- 
sidering its  surroundings:  in  some  languages  it  is  not  possible  to 
say  "sit  in  the  swn." 

Everything  said  about  words  apphes  to  any  term  of  language, 
e.g.  to  grammatical  entities.  The  value  of  a  French  plural  does  not 
coincide  with  that  of  a  Sanskrit  plural  even  though  their  sig- 
nification is  usually  identical ;  Sanskrit  has  three  numbers  instead 
of  two  {my  eyes,  my  ears,  my  arms,  my  legs,  etc.  are  dual)  ;*  it  would 
be  wrong  to  attribute  the  same  value  to  the  plural  in  Sanskrit  and 
in  French;  its  value  clearly  depends  on  what  is  outside  and  around 

If  words  stood  for  pre-existing  concepts,  they  would  all  have 
exact  equivalents  in  meaning  from  one  language  to  the  next;  but 
this  is  not  true.  French  uses  louer  (une  maison)  'let  (a  house)'  in- 
differently to  mean  both  "pay  for"  and  "receive  pajrment  for," 
whereas  German  uses  two  words,  mieten  and  vermieten;  there  is 
obviously  no  exact  correspondence  of  values.  The  German  verbs 
schdtzen  and  urteilen  share  a  number  of  significations,  but  that 
correspondence  does  not  hold  at  several  points. 

Inflection  offers  some  particularly  striking  examples.  Dis- 
tinctions of  time,  which  are  so  familiar  to  us,  are  unknown  in  cer- 
tain languages.  Hebrew  does  not  recognize  even  the  fundamental 

*  The  use  of  the  comparative  form  for  two  and  the  superlative  for  more  than 
two  in  EngUsh  (e.g.  viay  the  better  hoxer  win:  the  best  boxer  in  the  world) 
is  probably  a  remnant  of  the  old  distinction  between  the  dual  and  the  plural 
number.  [Tr.] 



distinctions  between  the  past,  present,  and  future.  Proto-Germanic 
has  no  special  form  for  the  future;  to  say  that  the  future  is  ex- 
pressed by  the  present  is  wrong,  for  the  value  of  the  present  is  not 
the  same  in  Germanic  as  in  languages  that  have  a  future  along  with 
the  present.  The  Slavic  languages  regularly  single  out  two  aspects 
of  the  verb :  the  perfective  represents  action  as  a  point,  complete  in 
its  totality;  the  imperfective  represents  it  as  taking  place,  and  on 
the  line  of  time.  The  categories  are  difficult  for  a  Frenchman  to 
understand,  for  they  are  unknown  in  French;  if  they  were  pre- 
determined, this  would  not  be  true.  Instead  of  pre-existing  ideas 
then,  we  find  in  all  the  foregoing  examples  values  emanating  froijci 
the  system.  When  they  are  said  to  correspond  to  concepts,  it  is  I 
understood  that  the  concepts  are  purely  differential  and  defined! 
not  by  their  positive  content  but  negatively  by  their  relations  with 
the  other  terms  of  the  system.  Their  most  precise  characteristic  Uj 
in  being  what  the  others  are  not. 

Now  the  real  interpretation  of  the  diagram  of  the  signal  becomes 
apparent.  Thus 

means  that  in  French  the  concept  "to  judge"  is-iinketd  to  the  soundU--^ 
imagQ..Juger;  in  short,  it  symbolizes  signiEGation.  But  it  is  quite 

•''ciear  that  ffiitiaiiy-^'e^oncept  is  nothing,  that  is  only  a  value 
determined  by  its  relations  with  other  similar  values,  and  that 

'^without  them  the  signification  would  not  exist.-  If  I  state  simply- 
that  a  word  signifies  somethiirg  wh^SrrTTiave  in  mind  the  associ- 
ating of  a  sound-image  with  a  concept,  I  am  making  a  statement 
that  may  suggest  what  actually  happens,  but  by  no  means  am  I 
expressing  the  linguistic  fact  in  its  essence  and  fullness. 

3.  Linguistic  Value  from  a  Material  Viewpoint 

The  conceptual  side  of  value  is  made  up  solely  of  relations  and 
differences  with  respect  to  the  other  terms  of  language,  and  the 


same  can  be  said  of  its  material  side.  The  important  thing  in  the 
word  is  not  the  sound  alone  but  the  phonic  differences  that  make 
it  possible  to  distinguish  this  word  from  all  others,  for  differences 
carry  signification. 

This  may  seem  surprising,  but  how  indeed  could  the  reverse  be 
possible?  Since  one  vocal  image  is  no  better  suited  than  the  next 
for  what  it  is  commissioned  to  express,  it  is  evident,  even  a  priori, 
that  a  segment  of  language  can  never  in  the  final  analysis  be  based 
on  anything  except  its  noncoincidence  with  the  rest.  Arbitrary  and 
differential  are  two  correlative  qualities. 

The  alteration  of  linguistic  signs  clearly  illustrates  this.  It  is 
precisely  because  the  terms  a  and  6  as  such  are  radically  incapable 
of  reaching  the  level  of  consciousness — one  is  always  conscious  of 
only  the  a/b  difference — that  each  term  is  free  to  change  accord- 
ing to  laws  that  are  unrelated  to  its  signifying  function.  No  positive 
sign  characterizes  the  genitive  plural  in  Czech  zen  (see  p.  86); 
still  the  two  forms  Sena:  zen  function  as  well  as  the  earlier  forms 
zena:  zenb;  zen  has  value  only  because  it  is  different. 

Here  is  another  example  that  shows  even  more  clearly  the  sys- 
tematic role  of  phonic  differences:  in  Greek,  ephen  is  an  imperfect 
and  esten  an  aorist  although  both  words  are  formed  in  the  same 
way;  the  first  belongs  to  the  system  of  the  present  indicative  of 
pheml  '1  say,'  whereas  there  is  no  present  *stem.i;  now  it  is  precisely 
the  relation  pheml:  ephen  that  corresponds  to  the  relation  between 
the  present  and  the  imperfect  (cf.  deiknumi:  edeiknun,  etc.).  Signs 
function,  then,  not  through  their  intrinsic  value  but  through  their 
relative  position. 

In  addition,  it  is  impossible  for  sound  alone,  a  material  element, 
to  belong  to  language.  It  is  only  a  secondary  thing,  substance  to  be 
put  to  use.  All  our  conventional  values  have  the  characteristic  of 
not  being  confused  with  the  tangible  element  which  supports  them. 
For  instance,  it  is  not  the  metal  in  a  piece  of  money  that  fixes  its 
value.  A  coin  nominally  worth  five  francs  may  contain  less  than 
half  its  worth  of  silver.  Its  value  will  vary  according  to  the  amount 
stamped  upon  it  and  according  to  its  use  inside  or  outside  a  politi- 
cal boundary.  This  is  even  more  true  of  the  linguistic  signifier, 
which  is  not  phonic  but  incorporeal — constituted  not  by  its  ma- 


terial  substance  but  by  the  differences  that  separate  its  sound- 
image  from  all  others. 

The  foregoing  principle  is  so  basic  that  it  applies  to  all  the 
material  elements  of  language,  including  phonemes.  Every  lan- 
guage forms  its  words  on  the  basis  of  a  system  of  sonorous  ele- 
ments, each  element  being  a  clearly  delimited  unit  and  one  of  a 
fixed  number  of  units.  Phonemes  are  characterized  not,  as  one 
might  think,  by  their  own  positive  quality  but  simply  by  the  fact 
that  they  are  distinct.  Phonemes  are  above  all  else  opposing, 
relative,  and  negative  entities. 

Proof  of  this  is  the  latitude  that  speakers  have  between  points 
of  convergence  in  the  pronunciation  of  distinct  sounds.  In  French, 
for  instance,  general  use  of  a  dorsal  r  does  not  prevent  many  speak- 
ers from  using  a  tongue-tip  trill;  language  is  not  in  the  least  dis- 
turbed by  it;  language  requires  only  that  the  sound  be  different 
and  not,  as  one  might  imagine,  that  it  have  an  invariable  quality. 
I  can  even  pronounce  the  French  r  like  German  ch  in  Bach,  dock, 
-etc.,  but  in  German  I  could  not  use  r  instead  of  ch,  for  German 
gives  recognition  to  both  elements  and  must  keep  them  apart. 
Similarly,  in  Russian  there  is  no  latitude  for  t  in  the  direction  of  t' 
(palatalized  t),  for  the  result  would  be  the  confusing  of  two  sounds 
differentiated  by  the  language  (cf.  govorit'  'speak'  and  goverit  'he 
speaks'),  but  more  freedom  may  be  taken  with  respect  to  th  (aspi- 
rated t)  since  this  sound  does  not  figure  in  the  Russian  system  of 

Si'nce  an  identical  state  of  affairs  is  observable  in  writing,  an- 
other system  of  signs,  we  shall  use  writing  to  draw  some  com- 
parisons that  will  clarify  the  whole  issue.  In  fact: 

1)  The  signs  used  in  writing  are  arbitrary;  there  is  no  con- 
nection, for  example,  between  the  letter  t  and  the  sound  that  it 

2)  The  value  of  letters  is  purely  negative  and  differential.  The 
same  person  can  write  /,  for  instance,  in  different  ways: 

^  -^ 



The  only  requirement  is  that  the  sign  for  t  not  be  confused  in  his 
script  with  the  signs  used  for  I,  d,  etc. 

3)  Values  in  writing  function  only  through  reciprocal  opposition 
within  a  fixed  system  that  consists  of  a  set  number  of  letters.  This 
third  characteristic,  though  not  identical  to  the  second,  is  closely 
related  to  it,  for  both  depend  on  the  first.  Since  the  graphic  sign  is 
arbitrary,  its  form  matters  little  or  rather  matters  only  within  the 
limitations  imposed  by  the  system. 

4)  The  means  by  which  the  sign  is  produced  is  completely  un- 
important, for  it  does  not  affect  the  system  (this  also  follows  from 
characteristic  1).  Whether  I  make  the  letters  in  white  or  black, 
raised  or  engraved,  with  pen  or  chisel — all  this  is  of  no  importance 
with  respect  to  their  signification. 

4.  The  Sign  Considered  in  Its  Totality 
r"^  Everything  that  has  been  said  up  to  this  poiut  boils  down-4;0 
\_JJbdsiin  language  there_are  only  differences.  (Even  more  importan1;?> 
\  a  difference  generally  implies  positive  terms  between  which  the  I 
1  difference  is  set  up;  but  in  language  there  are  only  differences  \ 
I  without  positive  terms.  Whether  we  take  the  signified  or  the  signifier,    I 
'language  has  neither  ideas  nor  sounds  that  existed  before  the  lin-    \ 
guistic  system,  but  only  conceptual  and  phonic,  differences  that 
have  issued  from  the  system.  The  idea  or  phonic  substance^hat  a 
"§igS~t;Dntain5Js  of  less  importance  than  the  other  signs  that  sur- 
round it.  Proof  of  this  is  that  the  value  of  a  term  may  be  modified 
j' without  either  its  meaning  or  its  sound  being  affected,  solely  be- 
i  cause  a  neighboring  term  has  been  modified  (see  p.  115). 

But  the  statement  that  everything  in  language  is  negative  is 
true  only  if  the  signified  and  the  signifier  are  considered  separately ; 
when  we  consider  the  sign  in  its  totality,  we  have  something  that 
is  positive  in  its  own  class.  A  linguistic  system  is  a  series  of  differ- 
ences of  sound  combined  with  a  series  of  differences  of  ideas;  but 
the  pairing  of  a  certain  number  of  acoustical  signs  with  as  many 
cuts  made  from  the  mass  of  thought  engenders  a  system  of  values ; 
and  this  system  serves  as  the  effective  hnk  between  the  phonic  and 
psychological  elements  within  each  sign.  Although  both  the  sig- 
nified and  the  signifier  are  purely  differential  and  negative  when 
considered  separately,  their  combination  is  a  positive  fact;  it  is 


even  the  sole  type  of  facts  that  language  has,  for  maintaining  the 
parallelism  between  the  two  classes  of  differences  is  the  distinctive 
function  of  the  linguistic  institution. 

Certain  diachronic  facts  are  typical  in  this  respect.  Take  the 
countless  instances  where  alteration  of  the  signifier  occasions  a 
conceptual  change  and  where  it  is  obvious  that  the  sum  of  the 
ideas  distinguished  corresponds  in  principle  to  the  sum  of  the  dis- 
tinctive signs.  When  two  words  are  confused  through  phonetic 
alteration  (e.g.  French  decrepit  from  decrepitus  and  decrepi  from 
crispus),  the  ideas  that  they  express  will  also  tend  to  become  con- 
fused if  only  they  have  something  in  common.  Or  a  word  may  have 
different  forms  (cf.  chaise  'chair'  and  chaire  'desk').  Any  nascent 
difference  will  tend  invariably  to  become  significant  but  without 
always  succeeding  or  being  successful  on  the  first  trial.  Conversely, 
any  conceptual  difference  perceived  by  the  mind  seeks  to  find  ex- 
pression through  a  distinct  signifier,  and  two  ideas  that  are  no 
longer  distinct  in  the  mind  tend  to  merge  into  the  same  signifier. 

When  we  compare  signs — positive  terms — with  each  other,  we 
can  no  longer  speak  of  difference;  the  expression  would  not  be 
fitting,  for  it  apphes  only  to  the  comparing  of  two  sound-images, 
e.g.  father  and  mother,  or  two  ideas,  e.g.  the  idea  "father"  and  the 
idea  "mother";  two  signs,  each  having  a  signified  and  signifier,  are 
not  different  but  only  distinct.  Between  them  there  is  only  oppo- 
sition. The  entire  mechanism  of  language,  with  which  we  shall  be 
concerned  later,  is  based  on  oppositions  of  this  kind  and  on  the 
phonic  and  conceptual  differences  that  they  imply. 

What  is  true  of  value  is  true  also  of  the  unit  (see  pp.  110  ff.).  A 
unit  is  a  segment  of  the  spoken  chain  that  corresponds  to  a  certain 
concept;  both  are  by  nature  purely  differential.  i 

Applied  to  units,  the  principle  of  differentiation  can  be  stated  in 
this  way :  the  characteristics  of  the  unit  blend  with  the  unit  itself.  In 
language,  as  in  any  semiological  system,  whatever  distinguishes 
one  sign  from  the  others  constitutes  it.  Difference  makes  character 
just  as  it  makes  value  and  the  unit. 

Another  rather  paradoxical  consequence  of  the  same  principle  is 
this:  in  the  last  analysis  what  is  commonly  referred  to  as  a  "gram- 
matical fact"  fits  the  definition  of  the  unit,  for  it  always  expresses 
an  opposition  of  terms;  it  differs  only  in  that  the  opposition  is 


particularly  significant  (e.g.  the  formation  of  German  plurals  of  the 
type  Nacht:  Ndchte).  Each  term  present  in  the  grammatical  fact 
(the  singular  without  umlaut  or  final  e  in  opposition  to  the  plural 
with  umlaut  and  -e)  consists  of  the  interplay  of  a  number  of  oppo- 
sitions within  the  system.  When  isolated,  neither  Nacht  nor  Ndchte 
is  anything:  thus  everything  is  opposition.  Putting  it  another  way, 
the  Nacht:  Ndchte  relation  can  be  expressed  by  an  algebraic  formula 
a/b  in  which  a  and  h  are  not  simple  terms  but  result  from  a  set  of 
relations.  Language,  in  a  manner  of  speaking,  is  a  type  of  algebra 
consisting  solely  of  complex  terms.  Some  of  its  oppositions  are  more 
significant  than  others;  but  units  and  grammatical  facts  are  only 
different  names  for  designating  diverse  aspects  of  the  same  general 
fact :  the  functioning  of  linguistic  oppositions.  This  statement  is  so 
true  that  we  might  very  well  approach  the  problem  of  units  by 
starting  from  grammatical  facts.  Taking  an  opposition  like  Nacht: 
Ndchte,  we  might  ask  what  are  the  units  involved  in  it.  Are  they 
only  the  two  words,  the  whole  series  of  similar  words,  a  and  d,  or  all 
singulars  and  plurals,  etc.? 

Units  and  grammatical  facts  would  not  be  confused  if  linguistic 
signs  were  made  up  of  something  besides  differences.  But  language 
being  what  it  is,  we  shall  find  nothing  simple  in  it  regardless  of  our 
approach;  everywhere  and  always  there  is  the  same  complex 
equilibrium  of  terms  that  mutually  condition  each  other.  Putting 
it  another  way,  language  is  a  form  and  not  a  substance  (see  p.  113). 
This  truth  could  not  be  overstressed,  for  all  the  mistakes  in  our 
terminology,  all  our  incorrect  ways  of  naming  things  that  pertain 
to  language,  stem  from  the  involuntary  supposition  that  the 
linguistic  phenomenon  must  have  substance. 

Chapter  V 


1.  Definitions 

In  a  language-state  everything  is  based  on  relations.  How  do 
they  function? 


Relations  and  differences  between  linguistic  terms  fall  into  two 
distinct  groups,  each  of  which  generates  a  certain  class  of  values. 
The  opposition  between  the  two  classes  gives  a  better  understand- 
ing of  the  nature  of  each  class.  They  correspond  to  two  forms  of 
our  mental  activity,  both  indispensable  to  the  life  of  language. 

In  discourse,  on  the  one  hand,  words  acquire  relations  based  on 
the  linear  nature  of  language  because  they  are  chained  together. 
This  rules  out  the  possibihty  of  pronouncing  two  elements  simul- 
taneously (see  p.  70).  The  elements  are  arranged  in  sequence  on 
the  chain  of  speaking.  Combinations  supported  by  linearity  are 
syntagms.^  The  syntagm  is  always  composed  of  two  or  more  con- 
secutive units  (e.g.  French  re-lire  're-read,'  contre  tous  'against 
everyone,'  la  vie  humaine  'human  life,'  Dieu  est  bon  'God  is  good,' 
s'il  fait  beau  temps,  nous  sortirons  'if  the  weather  is  nice,  we'll  go 
out,'  etc.).  In  the  syntagm  a  term  acquires  its  value  only  because 
it  stands  in  opposition  to  everything  that  precedes  or  follows  it, 
or  to  both. 

Outside  discourse,  on  the  other  hand,  words  acquire  relations  of 
a  different  kind.  Those  that  have  something  in  common  are  asso- 
ciated in  the  memory,  resulting  in  groups  marked  by  diverse  re- 
lations. For  instance,  the  French  word  enseignement  'teaching'  will 
unconsciously  call  to  mind  a  host  of  other  words  (enseigner  'teach,' 
renseigner  'acquaint,'  etc.;  or  armement  'armament,'  changement 
'amendment,'  etc.;  or  education  'education,'  apprentissage  'ap- 
prenticeship,' etc.).  All  those  words  are  related  in  some  way. 

We  see  that  the  co-ordinations  formed  outside  discourse  differ 
strikingly  from  those  formed  inside  discourse.  Those  formed  out- 
side discourse  are  not  supported  by  linearity.  Their  seat  is  in  the 
brain;  they  are  a  part  of  the  inner  storehouse  that  makes  up  the 
language  of  each  speaker.  They  are  associative  relations. 

The  syntagmatic  relation  is  in  praesentia.  It  is  based  on  two  or 
more  terms  that  occur  in  an  effective  series.  Against  this,  the  associ- 
ative relation  unites  terms  in  absentia  in  a  potential  mnemonic 

From  the  associative  and  syntagmatic  viewpoint  a  linguistic 

^  It  is  scarcely  necessary  to  point  out  that  the  study  of  syntagms  is  not  to  be 
confused  with  syntax.  Syntax  is  only  one  part  of  the  study  of  syntagms 
(see  pp.  134  ff.).  [Ed.] 


unit  is  like  a  fixed  part  of  a  building,  e.g.  a  column.  On  the  one 
hand,  the  column  has  a  certain  relation  to  the  architrave  that  it 
supports;  the  arrangement  of  the  two  units  in  space  suggests  the 
syntagmatic  relation.  On  the  other  hand,  if  the  column  is  Doric,  it 
suggests  a  mental  comparison  of  this  style  with  others  (Ionic, 
Corinthian,  etc.)  although  none  of  these  elements  is  present  in 
space:  the  relation  is  associative. 

Each  of  the  two  classes  of  co-ordination  calls  for  some  specific 

2.  Syntagmatic  Relations 

The  examples  on  page  123  have  already  indicated  that  the  notion 
of  syntagm  applies  not  only  to  words  but  to  groups  of  words,  to 
complex  units  of  all  lengths  and  types  (compounds,  derivatives, 
phrases,  whole  sentences). 

It  is  not  enough  to  consider  the  relation  that  ties  together  the 
different  parts  of  syntagms  (e.g.  French  contre  'against'  and  tons 
'everyone'  in  contre  tous,  contre  and  maitre  'master'  in  contremattre 
'foreman')  f  one  must  also  bear  in  mind  the  relation  that  links  the 
whole  to  its  parts  (e.g.  contre  tous  in  opposition  on  the  one  hand  to 
contre  and  on  the  other  tous,  or  contremattre  in  opposition  to  contre 
and  maitre). 

An  objection  might  be  raised  at  this  point.  The  sentence  is  the 
ideal  type  of  syntagm.  But  it  belongs  to  speaking,  not  to  language 
(see  p.  14).  Does  it  not  follow  that  the  syntagm  belongs  to  speak- 
ing? I  do  not  think  so.  Speaking  is  characterized  by  freedom 
of  combinations;  one  must  therefore  ask  whether  or  not  all  syn- 
tagms are  equally  free. 

It  is  obvious  from  the  first  that  many  expressions  belong  to  lan- 
guage. These  are  the  pat  phrases  in  which  any  change  is  prohibited 
by  usage,  even  if  we  can  single  out  their  meaningful  elements  (cf. 
a  quoi  bonf  'what's  the  use?'  allons  done!  'nonsense!').  The  same  is 
true,  though  to  a  lesser  degree,  of  expressions  like  prendre  la  mouche 
'take  offense  easily,'^  forcer  la  main  d  quelgii^un  'force  someone's 
hand,'  rompre  une  lance  'break  a  lance,'*  or  even  avoir  mal  (d  la 

*  Cf.  English  head  and  waiter  in  headwaiter.  [Tr.] 

^  Literally  'take  the  fly.'  Cf.  English  take  the  bull  by  the  horns.  [Tr.] 

«  Cf.  English  bury  the  hatchet.  [Tr.] 


tete,  etc.)  'have  (a  headache,  etc.),'  a  force  de  {soins,  etc.)  'by  dint  of 
(care,  etc.),'  que  vous  en  semblef  'how  do  you  feel  about  it?'  pas 
n'est  besoin  de  .  .  .  'there's  no  need  for  .  .  .,'  etc.,  which  are  charac- 
terized by  peculiarities  of  signification  or  syntax.  These  idiomatic 
twists  cannot  be  improvised;  they  are  furnished  by  tradition. 
There  are  also  words  which,  while  lending  themselves  perfectly  to 
analysis,  are  characterized  by  some  morphological  anomaly  that  is 
kept  solely  by  dint  of  usage  (cf .  difficulte  'difficulty'  beside  facilite 
'facility,'  etc.,  and  mourrai  '[I]  shall  die'  beside  dormirai  '[I]  shall 
sleep'). 9 

There  are  further  proofs.  To  language  rather  than  to  speaking 
belong  the  syntagmatic  types  that  are  built  upon  regular  forms. 
Indeed,  since  there  is  nothing  abstract  in  language,  the  types  exist 
only  if  language  has  registered  a  sufficient  number  of  specimens. 
When  a  word  like  indecorable  arises  in  speaking  (see  pp.  167  ff.),  its 
appearance  supposes  a  fixed  type,  and  this  type  is  in  turn  possible 
only  through  remembrance  of  a  sufficient  number  of  similar  words 
belonging  to  language  (impardonable  'unpardonable,'  intolerable 
'intolerable,'  infatigable  'indefatigable,'  etc.).  Exactly  the  same  is 
true  of  sentences  and  groups  of  words  built  upon  regular  patterns. 
Combinations  like  la  terre  iourne  'the  world  turns,'  que  vous  dit-ilf 
'what  does  he  say  to  you?'  etc.  correspond  to  general  types  that  are 
in  turn  supported  in  the  language  by  concrete  remembrances. 

But  we  must  realize  that  in  the  syntagm  there  is  no  clear-cut 
boundary  between  the  language  fact,  which  is  a  sign  of  collective 
usage,  and  the  fact  that  belongs  to  speaking  and  depends  on  indi- 
vidual freedom.  In  a  great  number  of  instances  it  is  hard  to  class  a 
combination  of  units  because  both  forces  have  combined  in  produc- 
ing it,  and  they  have  combined  in  indeterminable  proportions. 

3.  Associative  Relations 

Mental  association  creates  other  groups  besides  those  based  on 
the  comparing  of  terms  that  have  something  in  common ;  through 
its  grasp  of  the  nature  of  the  relations  that  bind  the  terms  together, 
the  mind  creates  as  many  associative  series  as  there  are  diverse 
relations.  For  instance,  in  enseignement  'teaching,'  enseigner  'teach,' 

'  The  anomaly  of  the  double  r  in  the  future  forms  of  certain  verbs  in  French 
may  be  compared  to  irregular  plurals  like  oxen  in  English.  [Tr.] 


enseignons  '(we)  teach,'  etc.,  one  element,  the  radical,  is  common 
to  every  term;  the  same  word  may  occur  in  a  different  series  formed 
around  another  common  element,  the  suffix  (cf .  enseignement,  arme- 
ment, changement,  etc.);  or  the  association  may  spring  from  the 
analogy  of  the  concepts  signified  {enseignement,  instruction,  ap- 
prentissage,  education,  etc.);  or  again,  simply  from  the  similarity 
of  the  sound-images  (e.g.  enseignement  and  justement  'precisely').^" 
Thus  there  is  at  times  a  double  similarity  of  meaning  and  form, 
at  times  similarity  only  of  form  or  of  meaning.  A  word  can  always 
evoke  everything  that  can  be  associated  with  it  in  one  way  or 

Whereas  a  syntagm  immediately  suggests  an  order  of  succession 
and  a  fixed  number  of  elements,  terms  in  an  associative  family 
occur  neither  in  fixed  numbers  nor  in  a  definite  order.  If  we  associ- 
ate painful,  delightful,  frightfid,  etc.  w^e  are  unable  to  predict  the 
number  of  words  that  the  memory  will  suggest  or  the  order  in 
which  they  will  appear.  A  particular  word  is  like  the  center  of  a 
constellation ;  it  is  the  point  of  convergence  of  an  indefinite  number 
of  co-ordinated  terms  (see  the  illustration  on  page  127). 

But  of  the  two  characteristics  of  the  associative  series — in- 
determinate order  and  indefinite  number — only  the  first  can  always 
be  verified ;  the  second  may  fail  to  meet  the  test.  This  happens  in 
the  case  of  inflectional  paradigms,  which  are  typical  of  associative 
groupings.  Latin  dominus,  dominl,  domino,  etc.  is  obviously  an 
associative  group  formed  around  a  common  element,  the  noun 
theme  domin-,  but  the  series" 

[  enseig 


ment  j 

enseigner         / 

\        clement 

enseignons            / 
©tc.            apprentissage 

\           juste 






"  The  last  case  is  rare  and  can  be  classed  as  abnormal,  for  the  mind  naturally 


is  not  indefinite  as  in  the  case  of  enseignement,  changement,  etc. ;  the 
number  of  cases  is  definite.  Against  this,  the  words  have  no  fixed 
order  of  succession,  and  it  is  by  a  purely  arbitrary  act  that  the 
grammarian  groups  them  in  one  way  rather  than  in  another;  in  the 
mind  of  speakers  the  nominative  case  is  by  no  means  the  first  one 
in  the  declension,  and  the  order  in  which  terms  are  called  depends 
on  circumstances. 

Chapter  VI 

1.  Syntagmatic  Solidarities 

The  set  of  phonic  and  conceptual  differences  that  constitutes 
language  results  from  two  types  of  comparisons;  the  relations  are 
sometimes  associative,  sometimes  syntagmatic.  The  groupings  in 
both  classes  are  for  the  most  part  fixed  by  language;  this  set  of 
common  relations  constitutes  language  and  governs  its  functioning. 

What  is  most  striking  in  the  organization  of  language  are  syntag- 
matic solidarities;  almost  all  units  of  language  depend  on  what 
surrounds  them  in  the  spoken  chain  or  on  their  successive  parts. 

This  is  shown  by  word  formation.  A  unit  like  painful  decomposes 

discards  associations  that  becloud  the  intelligibility  of  discourse.  But  its 
existence  is  proved  by  a  lower  category  of  puns  based  on  the  ridiculous  con- 
fusions that  can  result  from  pure  and  simple  homonomy  like  the  French 
statement:  "Les  musiciens  produisent  les  sons  ['sounds,  bran']  et  les  grainetiers 
les  vendent"  'musicians  produce  sons  and  seedsmen  sell  them.'  [Cf.  Shake- 
speare's "Not  on  thy  sole,  but  on  thy  soul."  (Tr.)]  This  is  distinct  from  the  case 
where  an  association,  while  fortuitous,  is  supported  by  a  comparison  of  ideas 
(cf.  French  ergot  'spur':  ergotcr  'wrangle';  German  blau  'blue':  durchblauen 
'thrash  soundly');  the  point  is  that  one  member  of  the  pair  has  a  new  in- 
terpretation. Folk  etymologies  like  these  (see  pp.  173  ff.)  are  of  interest  in  the 
study  of  semantic  evolution,  but  from  the  synchronic  viewpoint  they  are  in 
the  same  category  as  enseigner:  enseignement.  [Ed.] 

"  Cf.  Enghsh  education  and  the  corresponding  associative  series:  educate, 
educates,  etc.;  internship,  training,  etc.;  vocation,  devotion,  etc.;  and  lotion, 
fashion,  etc.  [Tr.] 


into  two  subunits  (pain-ful),  but  these  subunits  are  not  two  inde- 
pendent parts  that  are  simply  lumped  together  (pain  +  ful) .  The 
unit  is  a  product,  a  combination  of  two  interdependent  elements 
that  acquire  value  only  through  their  reciprocal  action  in  a  higher 
unit  (pain  X  ful).  The  suffix  is  nonexistent  when  considered  inde- 
pendently; what  gives  it  a  place  in  the  language  is  a  series  of  com- 
mon terms  like  delight-ful,  fright-ful,  etc.  Nor  is  the  radical  inde- 
pendent. It  exists  only  through  combining  with  a  suffix.  In  gos-Ung, 
the  element  gos-  is  nothing  without  its  sufiix.  The  whole  has  value 
only  through  its  parts,  and  the  parts  have  value  by  virtue  of  their 
place  in  the  whole.  That  is  why  the  syntagmatic  relation  of  the  part 
to  the  whole  is  just  as  important  as  the  relation  of  the  parts  to  each 

This  general  principle  holds  true  for  every  type  of  syntagm 
enumerated  above  (pp.  124  ff.),  for  larger  units  are  always  com- 
posed of  more  restricted  units  linked  by  their  reciprocal  solidarity. 

To  be  sure,  language  has  independent  units  that  have  syntag- 
matic relations  with  neither  their  parts  nor  other  units.  Sentence 
equivalents  like  yes,  no,  thanks,  etc.  are  good  examples.  But  this 
exceptional  fact  does  not  compromise  the  general  principle.  As  a 
rule  we  do  not  communicate  through  isolated  signs  but  rather 
through  groups  of  signs,  through  organized  masses  that  are  them- 
selves signs.  In  language  everything  boils  down  to  differences  but 
also  to  groupings.  The  mechanism  of  language,  which  consists  of 
the  interplay  of  successive  terms,  resembles  the  operation  of  a 
machine  in  which  the  parts  have  a  reciprocating  function  even 
though  they  are  arranged  in  a  single  dimension. 

2.  Simultaneous  Functioning  of  the  Two  Types  of  Groupings 

Between  the  sj^ntagmatic  groupings,  as  defined,  there  is  a  bond 
of  interdependence;  they  mutually  condition  each  other.  In  fact, 
spatial  co-ordinations  help  to  create  associative  co-ordinations, 
which  are  in  turn  necessary  for  analysis  of  the  parts  of  the  syntagm. 
Take  the  French  compound  de-faire  'un-do.'  ^^  We  can  picture  it 
as  a  horizontal  ribbon  that  corresponds  to  the  spoken  chain: 

12  Cf.  English  misplace.  To  the  French  series  correspond  English  mistake, 
misspell,  misrepresent,  etc.  and  place,  replace,  displace,  etc.  [Tr.] 



But  simultaneously  and  on  another  axis  there  exists  in  the  sub- 
conscious one  or  more  associative  series  comprising  units  that  have 
an  element  in  common  with  the  syntagm : 












In  the  same  way,  if  Latin  quadrwplex  is  a  syntagm,  this  is  because  it 
too  is  supported  by  a  double  associative  series : 










To  the  extent  that  the  other  forms  float  around  defaire  or  quadru- 
plex,  these  words  can  be  decomposed  into  subunits.  This  is  just  an- 
other way  of  saying  that  they  are  syntagms.  Defaire  could  not  be 
analyzed,  for  instance,  if  the  other  forms  containing  de-  or  faire 
disappeared  from  the  language.  It  would  be  but  a  simple  unit,  and 
its  two  parts  could  not  be  placed  in  opposition. 

Now  the  functioning  of  the  dual  system  in  discourse  is  clear. 


Our  memory  holds  in  reserve  all  the  more  or  less  complex  types 
of  syntagms,  regardless  of  their  class  or  length,  and  we  bring  in  the 
associative  groups  to  fix  our  choice  when  the  time  for  using  them 
arrives.  When  a  Frenchman  says  marchons!  '(let's)  walk!'  he 
thinks  unconsciously  of  diverse  groups  of  associations  that  con- 
verge on  the  syntagm  marchons!  The  syntagm  figures  in  the  series 
marche!  '(thou)  walk!'  marchez!  '(you)  walk!'  and  the  opposition 
between  marchons!  and  the  other  forms  determines  his  choice;  in 
addition,  marchons!  calls  up  the  series  montons!  '(let's)  go  up!' 
mangeons  '(let's)  eat!'  etc.  and  is  selected  from  the  series  by  the 
same  process.  In  each  series  the  speaker  knows  what  he  must  vary 
in  order  to  produce  the  differentiation  that  fits  the  desired  unit.  If 
he  changes  the  idea  to  be  expressed,  he  will  need  other  oppositions 
to  bring  out  another  value;  for  instance,  he  may  say  marchez!  or 
perhaps  montons! 

It  is  not  enough  to  say,  looking  at  the  matter  positively,  that  the 
speaker  chooses  marchons!  because  it  signifies  what  he  wishes  to 
express.  In  reality  the  idea  evokes  not  a  form  but  a  whole  latent 
system  that  makes  possible  the  oppositions  necessary  for  the  for- 
mation of  the  sign.  By  itself  the  sign  would  have  no  signification. 
If  there  were  no  forms  like  marche!  marchez!  against  marchons!, 
certain  oppositions  w^ould  disappear,  and  the  value  of  marchons! 
would  be  changed  ipso  facto. 

This  principle  applies  to  even  the  most  complex  types  of  syn- 
tagms and  sentences.  To  frame  the  question  que  vous  dit-il?  'what 
does  he  say  to  youf  the  speaker  varies  one  element  of  a  latent 
syntactical  pattern,  e.g.  que  te  dit-il?  'what  does  he  say  to  theef^ 
que  nous  dit-il?  'what  does  he  say  to  usf  etc.,  until  his  choice  is 
fixed  on  the  pronoun  vous.  In  this  process,  which  consists  of  elimi- 
nating mentally  everything  that  does  not  help  to  bring  out  the 
desired  differentiation  at  the  desired  point,  associative  groupings 
and  sjmtagmatic  patterns  both  play  a  role. 

Conversely,  the  process  of  fixation  and  choice  governs  the 
smallest  units  and  even  phonological  elements  wherever  they  are 
endowed  with  a  value.  I  am  thinking  not  only  of  cases  like  French 
pQtit  'small'  (feminine  form,  written  petite)  in  opposition  to  p9ti 
(masculine  form,  written  petit)  or  Latin  domini  against  domino, 
where  the  difference  happens  to  be  based  on  a  simple  phoneme,  but 


also  of  the  more  subtle  and  characteristic  fact  that  a  phoneme  by 
itself  plays  a  role  in  the  system  of  a  language-state.  For  example,  if 
m,  p,  t,  etc.  can  never  occur  at  the  end  of  a  word  in  Greek,  this 
means  that  their  presence  or  absence  in  a  definite  position  counts 
in  the  structure  of  the  word  and  in  the  structure  of  the  sentence. 
In  every  such  case  the  isolated  sound,  like  every  other  unit,  is 
chosen  after  a  dual  mental  opposition.  In  the  imaginary  grouping 
anma,  for  instance,  the  sound  m  stands  in  syntagmatic  opposition 
to  its  environing  sounds  and  in  associative  opposition  to  all  other 
sounds  that  may  come  to  mind: 




3.  Absolute  and  Relative  Arbitrariness 

The  mechanism  of  language  can  be  presented  from  another 
especially  important  angle. 

The  fundamental  principle  of  the  arbitrariness  of  the  sign  does 
not  prevent  our  singling  out  in  each  language  what  is  radically 
arbitrary,  i.e.  unmotivated,  and  what  is  only  relatively  arbitrary. 
Some  signs  are  absolutely  arbitrary;  in  others  we  note,  not  its  com- 
plete absence,  but  the  presence  of  degrees  of  arbitrariness :  the  sign 
may  be  relatively  motivated. 

For  instance,  both  vingt  'twenty'  and  dix-neuf  'nineteen'  are  un- 
motivated in  French,  but  not  in  the  same  degree,  for  dix-neuf 
suggests  its  own  terms  and  other  terms  associated  with  it  (e.g.  dix 
'ten,'  neuf  'nine,'  vingtr-neuf  'twenty-nine,'  dix-huit  'eighteen,' 
soixante-dix  'seventy,'  etc.).  Taken  separately,  dix  and  neuf  are  in 
the  same  class  as  vingt,  but  dix-neuf  is  an  example  of  relative  mo- 
tivation. The  same  is  true  of  poirier  'pear-tree,'  which  recalls  the 
simple  word  poire  'pear'  and,  through  its  suffix,  cerisier  'cherry- 
tree,'  pommier  'apple-tree,'  etc.^'  For  fr^ne  'ash,'  ch^ne  'oak,'  etc. 
there  is  nothing  comparable.  Again,  compare  berger  'shepherd,' 
which  is  completely  unmotivated,  and  vacher  'cowherd,'  which  is 
relatively  motivated.^*  In  the  same  way,  the  pairs  gedle  'jail'  and 

"  Cf.  English  flaxen,  which  suggests  flax,  silken,  woolen,  etc.  [Tr.] 

"  Cf.  English  clerk,  unmotivated,  against /armer,  relatively  motivated.  [Tr.] 


cachot  'dungeon,'  hache  *ax'  and  couperet  'chopper/  concierge  'por- 
ter' and  portier  'doorman,'  jadis  'of  old'  and  autrefois  'formerly,' 
souvent  'often'  and  frequemment  'frequently,'  aveugle  'blind'  and 
boiteux  'limping,'  sourd  'deaf  and  bossu  'hunchbacked,'  second 
'second'  and  deuxieme  'second  (of  a  series),'  German  Laub  and 
French,  feuillage  'foliage,'  and  French  metier  'handicraft'  and  Ger- 
man Handwerk.^^  The  English  plural  ships  suggests  through  its 
formation  the  whole  series  flags,  birds,  books,  etc.,  while  men  and 
sheep  suggest  nothing.  In  Greek  doso  'I  shall  give'  the  notion  of 
futurity  is  expressed  by  a  sign  that  calls  up  the  association  luso, 
steso,  tupso,  etc.;  eimi  'I  shall  go,'  on  the  other  hand,  is  completely 

This  is  not  the  place  to  search  for  the  forces  that  condition 
motivation  in  each  instance;  but  motivation  varies,  being  always 
proportional  to  the  ease  of  syntagmatic  analysis  and  the  obvious- 
ness of  the  meaning  of  the  subunits  present.  Indeed,  while  some 
formative  elements  hke  -ier  in  poir-ier  against  ceris-ier,  pomm-ier, 
etc.  are  obvious,  others  are  vague  or  meaningless.  For  instance, 
does  the  sufi&x  -ot  really  correspond  to  a  meaningful  element  in 
French  cachot  'dungeon'?  On  comparing  words  like  coutelas  'cutlas,' 
fatras  'pile,'  platras  'rubbish,'  canevas  'canvas,'  etc.,  one  has  no 
more  than  the  vague  feeUng  that  -as  is  a  formative  element  charac- 
teristic of  substantives.  At  any  rate,  even  in  the  most  favorable 
cases  motivation  is  never  absolute.  Not  only  are  the  elements  of  a 
motivated  sign  themselves  unmotivated  (cf.  dix  and  neuf  in  dix- 
neuf),  but  the  value  of  the  whole  term  is  never  equal  to  the  sum  of 
the  value  of  the  parts.  Teach  +  er  is  not  equal  to  teach  X  er  (see 
p.  128). 

Motivation  is  explained  by  the  principles  stated  in  Section  2. 
The  notion  of  relative  motivation  impUes:  (1)  analysis  of  a  given 
term,  hence  a  syntagmatic  relation ;  and  (2)  the  summoning  of  one 
or  more  other  terms,  hence  an  associative  relation.  It  is  the 
mechanism  through  which  any  term  whatever  lends  itself  to  the 
expression  of  an  idea,  and  is  no  more  than  that.  Up  to  this  point 
units  have  appeared  as  values,  i.e.  as  elements  of  a  system,  and  we 

1^  For  examples  not  similar  in  English  and  French,  compare  completely 
unmotivated  jail,  slave,  then  and  relatively  motivated  reformatory,  servant, 
heretofore.  [Tr.] 


have  given  special  consideration  to  their  opposition ;  now  we  recog- 
nize the  solidarities  that  bind  them;  they  are  associative  and 
syntagmatic,  and  they  are  what  limits  arbitrariness.  Dix-neuf  is 
supported  associatively  by  dix-huit,  soixante-dix,  etc.  and  syntag- 
matically  by  its  elements  dix  and  neuf  (see  p.  128).  This  dual 
relation  gives  it  a  part  of  its  value. 

Everything  that  relates  to  language  as  a  system  must,  I  am  con- 
vinced, be  approached  from  this  viewpoint,  which  has  scarcely 
received  the  attention  of  linguists:  the  limiting  of  arbitrariness. 
This  is  the  best  possible  basis  for  approaching  the  study  of  language 
as  a  system.  In  fact,  the  whole  system  of  language  is  based  on  the 
irrational  principle  of  the  arbitrariness  of  the  sign,  which  would 
lead  to  the  worst  sort  of  complication  if  applied  without  restriction. 
But  the  mind  contrives  to  introduce  a  principle  of  order  and  regu- 
larity into  certain  parts  of  the  mass  of  signs,  and  this  is  the  role  of 
relative  motivation.  If  the  mechanism  of  language  were  entirely 
rational,  it  could  be  studied  independently.  Since  the  mechanism 
of  language  is  but  a  partial  correction  of  a  system  that  is  by  nature 
chaotic,  however,  we  adopt  the  viewpoint  imposed  by  the  very 
nature  of  language  and  study  it  as  it  limits  arbitrariness. 

There  is  no  language  in  which  nothing  is  motivated,  and  our 
definition  makes  it  impossible  to  conceive  of  a  language  in  which 
everything  is  motivated.  Between  the  two  extremes — a  minimum 
of  organization  and  a  minimum  of  arbitrariness — we  find  all  pos- 
sible varieties.  Diverse  languages  always  include  elements  of  both 
types — radically  arbitrary  and  relatively  motivated — but  in  pro- 
portions that  vary  greatly,  and  this  is  an  important  characteristic 
that  may  help  in  classifying  them. 

In  a  certain  sense — one  which  must  not  be  pushed  too  far  but 
which  brings  out  a  particular  form  that  the  opposition  may  take — 
we  might  say  that  languages  in  which  there  is  least  motivation  are 
more  lexicological,  and  those  in  which  it  is  greatest  are  more  gram- 
matical. Not  because  "lexical"  and  ''arbitrary"  on  the  one  hand 
and  "grammar"  and  "relative  motivation"  on  the  other,  are  always 
synonymous,  but  because  they  have  a  common  principle.  The  two 
extremes  are  like  two  poles  between  which  the  whole  system  moves, 
two  opposing  currents  which  share  the  movement  of  language :  the 
tendency  to  use  the  lexicological  instrument  (the  unmotivated 


sign)  and  the  preference  given  to  the  grammatical  instrument 
(structural  rules). 

We  would  see,  for  example,  that  motivation  plays  a  much  larger 
role  in  German  than  in  English.  But  the  ultra-lexicological  type  is 
Chinese  while  Proto-Indo-European  and  Sanskrit  are  specimens  of 
the  ultra^grammatical  type.  Within  a  given  language,  all  evolution- 
ary movement  may  be  characterized  by  continual  passage  from 
motivation  to  arbitrariness  and  from  arbitrariness  to  motivation; 
this  see-saw  motion  often  results  in  a  perceptible  change  in  the 
proportions  of  the  two  classes  of  signs.  Thus  with  respect  to  Latin, 
French  is  characterized,  among  other  things,  by  a  huge  increase  in 
arbitrariness.  Latin  inimicus  recalls  in-  and  amicus  and  is  moti- 
vated by  them;  against  this,  ennemi  'enemy'  is  motivated  by 
nothing — it  has  reverted  1  o  absolute  arbitrariness,  which  is  really 
the  prime  characteristic  cf  the  linguistic  sign.  We  would  notice 
this  shift  in  hundreds  of  instances:  cf.  constdre  {stare):  couter 
'cost,'  fahrica  (faber):  forge  'forge,'  magister  (magis):  maltre 
'master,'  herhicarius  (herhix):  herger  'shepherd,'  etc.  French  owes 
its  characteristic  appearance  to  this  fact. 

Chapter  VII 

L  Definitions:  Traditional  Divisions 

Static  linguistics  or  the  description  of  a  language-state  is  gram- 
mar in  the  very  precise,  and  moreover  usual,  sense  that  the  word 
has  in  the  expressions  "grammar  of  the  Stock  Exchange,"  etc., 
where  it  is  a  question  of  a  complex  and  systematic  object  governing 
the  interplay  of  coexisting  values. 

Grammar  studies  language  as  a  system  of  means  of  expression. 
Grammatical  means  synchronic  and  significant,  and  since  no  sys- 
tem straddles  several  periods,  there  is  no  such  thing  as  "historical 
grammar";  the  discipline  so  labeled  is  really  only  diachronic 


My  definition  disagrees  with  the  narrower  one  usually  given. 
Morphology  and  syntax  together  are  what  is  generally  called  gram- 
mar while  lexicology,  or  the  science  of  words,  is  excluded. 

But  first,  do  these  divisions  fit  the  facts?  Do  they  agree  with  the 
principles  that  have  just  been  posited? 

Morphology  deals  with  different  classes  of  words  (verbs,  nouns, 
adjectives,  pronouns,  etc.)  and  with  different  inflectional  forms 
(conjugation,  declension,  etc.).  To  separate  this  study  from  syntax, 
it  is  alleged  that  syntax  has  as  its  object  the  functions  attached  to 
linguistic  units  while  morphology  considers  only  their  form.  For 
instance,  morphology  says  simply  that  the  genitive  of  Greek  phulax 
'guardian'  is  phulakos,  and  syntax  explains  the  use  of  the  two 

But  the  distinction  is  illusory.  The  series  of  forms  of  the  sub- 
stantive phulax  becomes  an  inflectional  paradigm  only  through 
comparison  of  the  functions  attached  to  the  different  forms; 
reciprocally,  the  functions  are  morphological  only  if  each  function 
corresponds  to  a  definite  phonic  sign.  A  declension  is  neither  a  Ust 
of  forms  nor  a  series  of  logical  abstractions  but  a  combination  of 
the  two  (see  pp.  102  £f.).  Forms  and  functions  are  interdependent 
and  it  is  difiicult,  if  not  impossible,  to  separate  them.  Linguistically, 
morphology  has  no  real,  autonomous  object.  It  cannot  form  a 
discipline  distinct  from  syntax. 

Second,  it  is  not  logical  to  exclude  lexicology  from  grammar.  As 
they  are  registered  in  the  dictionary,  words  do  not  seem  at  first 
glance  to  lend  themselves  to  grammatical  study,  which  is  generally 
restricted  to  the  relations  between  units.  But  we  notice  at  once 
that  innumerable  relations  may  be  expressed  as  eflEiciently  by 
words  as  by  grammar.  For  instance,  Latin  fid  and  facio  stand  in 
opposition  to  each  other  in  the  same  way  as  dlcor  and  died,  two 
grammatical  forms  of  the  same  word.  The  distinction  between  the 
perfective  and  imperfective  is  expressed  grammatically  in  Russian 
sprosit':  sprdsivat'  'ask'  and  lexicologically  in  skazdt':  govorit'  'say.' 
Prepositions  are  usually  assigned  to  grammar,  but  the  prepositional 
locution  en  consideration  de  'in  consideration  of  is  basically  lexi- 
cological since  the  word  consideration  retains  its  own  meaning  in 
the  French  phrase.  If  we  compare  Greek  peitho:  peithomai  with 
French  je  persuade  'I  persuade':  foheis  'I  obey,'  we  see  that  the 


opposition  is  expressed  grammatically  in  the  first  instance  and 
lexicologically  in  the  second.  A  large  number  of  relations  that  are 
expressed  in  certain  languages  by  cases  or  prepositions  are  rendered 
in  others  by  compounds,  more  like  words  proper  (French  royaume 
des  deux  'kingdom  of  heaven'  and  German  Himmelreich),  or  by 
derivatives  (French  moulin  d,  vent  'windmill'  and  Pohsh  wiatr-ak) 
or  finally,  by  simple  words  (French  hois  de  chauffage  'firewood'  and 
Russian  drovd,  French  bois  de  construction  'timber'  and  Russian 
Us).  The  interchange  of  simple  words  and  phrases  within  the  same 
language  also  occurs  very  frequently  (cf.  French  considerer  'con- 
sider' and  prendre  en  consideration  'take  into  consideration,'  se 
venger  de  'avenge'  and  tirer  vengeance  de  'take  revenge  on'). 

Functionally,  therefore,  the  lexical  and  the  syntactical  may 
blend.  There  is  basically  no  distinction  between  any  word  that  is 
not  a  simple,  irreducible  unit  and  a  phrase,  which  is  a  syntactical 
fact.  The  arrangement  of  the  subunits  of  the  word  obeys  the  same 
fundamental  principles  as  the  arrangement  of  groups  of  words  in 

In  short,  although  the  traditional  divisions  of  grammar  may  be 
useful  in  practice,  they  do  not  correspond  to  natural  distinctions. 
To  build  a  grammar,  we  must  look  for  a  different  and  a  higher 

2.  Rational  Divisions 

Morphology,  syntax,  and  lexicology  interpenetrate  because 
every  synchronic  fact  is  identical.  No  line  of  demarcation  can  be 
drawn  in  advance.  Only  the  distinction  established  above  between 
S3nitagmatic  and  associative  relations  can  provide  a  classification 
that  is  not  imposed  from  the  outside.  No  other  base  will  serve  for 
the  grammatical  system. 

We  should  first  gather  together  all  that  makes  up  a  language- 
state  and  fit  this  into  a  theory  of  syntagms  and  a  theory  of  associ- 
ations. Immediately  certain  parts  of  traditional  grammar  would 
seem  to  fall  effortlessly  into  one  category  or  the  other.  Inflection 
is  evidently  a  typical  kind  of  association  of  forms  in  the  mind  of 
speakers;  and  syntax  (i.e.  the  theory  of  word  groupings,  according 
to  the  most  common  definition)  goes  back  to  the  theory  of  syn- 
tagms, for  the  groupings  always  suppose  at  least  two  units  dis- 


tributed  in  space.  Not  every  syntagmatic  fact  is  classed  as  syn- 
tactical, but  every  syntactical  fact  belongs  to  the  syntagmatic 

To  prove  the  necessity  of  the  dual  approach,  almost  any  point 
of  grammar  will  do.  The  notion  of  word,  for  instance,  poses  two 
distinct  problems,  depending  on  whether  the  word  is  studied  from 
the  associative  or  the  syntagmatic  viewpoint.  In  French,  the 
adjective  grand  'big'  offers  a  duality  of  form  from  the  syntagmatic 
viewpoint  (grd  gargon  written  grand  gargon  'big  boy'  and  grat  dfa, 
written  grand  enfant  'big  baby')  and  another  duality  from  the 
associative  viewpoint  (masculine  gra,  written  grand,  and  feminine 
grad,  written  grande). 

Each  fact  should  in  this  way  be  fitted  into  its  syntagmatic  or 
associative  class,  and  the  whole  subject  matter  of  grammar  should 
be  arranged  along  its  two  natural  co-ordinates;  no  other  division 
will  show  what  must  be  changed  in  the  usual  framework  of  syn- 
chronic linguistics.  I  cannot  undertake  that  task  here,  for  my  aim 
is  limited  to  stating  only  the  most  general  principles. 

Chapter  VIII 


One  important  subject,  not  yet  touched  upon,  points  up  this  very 
necessity  of  examining  every  grammatical  question  from  the  two 
viewpoints  specified  in  Chapter  VII :  abstract  entities  in  grammar. 
Let  us  consider  them  first  associatively. 

To  associate  two  forms  is  not  only  to  feel  that  they  have  some- 
thing in  common  but  also  to  single  out  the  nature  of  the  relations 
that  govern  associations.  For  instance,  speakers  are  aware  that  the 
relation  between  enseigner  and  enseignement  or  juger  and  jugement 
is  not  the  same  as  the  relation  between  enseignement  and  jugement 
'judgment'  (see  p.  125).  This  is  how  the  system  of  associations 
is  tied  to  the  system  of  grammar.  We  can  say  that  the  sum  of  the 
conscious  and  methodical  classifications  made  by  the  grammarian 


who  studies  a  language-state  without  bringing  in  history  must 
coincide  with  the  associations,  conscious  or  not,  that  are  set  up  in 
speaking.  These  associations  fix  word-famiUes,  inflectional  para- 
digms, and  formative  elements  (radicals,  suflSxes,  inflectional  end- 
ings, etc.)  in  our  minds  (see  pp.  185  ff.). 

But  does  association  single  out  only  material  elements?  No,  of 
course  not.  We  have  already  seen  that  it  brings  together  words 
that  are  related  only  through  meaning  (cf.  enseignement,  ap- 
prentissage,  education,  etc.).  The  same  must  apply  in  grammar. 
Take  the  three  Latin  genitive  forms  domin-i,  reg-is,  ros-drum.  The 
sounds  of  the  three  endings  offer  no  basis  for  association,  yet  the 
endings  are  connected  by  the  feeling  that  they  have  a  common 
value  which  prescribes  an  identical  function.  This  suffices  to  create 
the  association  in  the  absence  of  any  material  support,  and  the 
notion  of  the  genitive  in  this  way  takes  its  place  in  the  language. 
Through  a  similar  procedure,  the  inflectional  endings  -ws,  -^,  -o, 
etc.  (in  dominus,  dominl,  domino,  etc.)  are  Unked  together  in  the 
mind  and  are  the  basis  for  the  more  general  notions  of  case  and  case 
endings.  Associations  of  the  same  class,  but  larger  still,  combine 
all  substantives,  adjectives,  etc.  and  fix  the  notion  of  parts  of 

All  these  things  exist  in  language,  but  as  abstract  entities;  their 
study  is  difficult  because  we  never  know  exactly  whether  or  not  the 
awareness  of  speakers  goes  as  far  as  the  analyses  of  the  gram- 
marian. But  the  important  thing  is  that  abstract  entities  are  always 
based,  in  the  last  analysis,  on  concrete  entities.  No  grammatical 
abstraction  is  possible  without  a  series  of  material  elements  as  a 
basis,  and  in  the  end  we  must  always  come  back  to  these  elements. 

Now  we  turn  to  the  syntagmatic  viewpoint.  The  value  of  a 
cluster  is  often  linked  to  the  order  of  its  elements.  In  analyzing  a 
syntagm,  the  speaker  does  not  restrict  himself  to  singling  out  its 
parts;  he  observes  a  certain  order  of  succession  among  them.  The 
meaning  of  English  pain-fid  or  Latin  signi-fer  depends  on  the 
respective  positions  of  their  subunits:  we  cannot  say  ful-pain  or 
fer-signum.  A  value  may  have  no  relations  with  a  concrete  element 
(hke  -ful  or  -fer)  and  result  solely  from  the  arrangement  of  the 
terms;  for  instance,  the  different  significations  of  the  two  clusters 
in  French  je  dois  'I  must'  and  dois-je?  'must  I?'  are  due  only  to 


word  order.  One  language  sometimes  expresses  through  word  or- 
der an  idea  that  another  would  convey  through  one  or  more  con- 
crete terms.  In  the  syntagmatic  pattern  gooseberry  wine,  gold  watch, 
etc.,  English  expresses  through  the  mere  order  of  the  terms  re- 
lations that  are  denoted  in  Modern  French  by  prepositions  (cf.  vin 
de  groseilles,  montre  en  or,  etc.).  Modern  French  in  turn  expresses 
the  notion  of  direct  complement  solely  through  putting  the  sub- 
stantive after  the  transitive  verb  (cf.  je  cueille  une  fleur  'I  pick  a 
flower'),  while  Latin  and  some  other  languages  use  the  accusative, 
which  is  characterized  by  special  case  endings,  etc. 

Word  order  is  unquestionably  an  abstract  entity,  but  it  owes  its 
existence  solely  to  the  concrete  units  that  contain  it  and  that  flow 
in  a  single  dimension.  To  think  that  there  is  an  incorporeal  syntax 
outside  material  units  distributed  in  space  would  be  a  mistake.  In 
English,  the  man  I  have  seen  apparently  uses  a  zero-sign  to  stand  for 
a  syntactical  fact  which  French  expresses  by  que  'that'  (I'homme 
que  j'ai  vu).  But  the  comparing  of  the  English  with  the  French 
syntactical  fact  is  precisely  what  produces  the  illusion  that 
nothingness  can  express  something.  The  material  units  alone 
actually  create  the  value  by  being  arranged  in  a  certain  way.  We 
cannot  study  a  syntactical  value  outside  a  number  of  concrete 
terms,  and  the  very  fact  that  we  understand  a  linguistic  complex 
(e.g.  the  English  words  cited  above)  shows  that  word-order  alone 
expresses  the  thought. 

A  material  unit  exists  only  through  its  meaning  and  function. 
This  principle  is  especially  important  in  understanding  smaller 
units,  for  one  is  tempted  to  think  that  they  exist  by  virtue  of  their 
sheer  material  quality — that  love,  for  example,  owes  its  existence 
solely  to  its  sounds.  Conversely — as  we  have  just  seen — a  meaning 
and  function  exist  only  through  the  support  of  some  material  form. 
This  principle  was  formulated  with  respect  to  larger  syntagms  or 
syntactical  patterns,  but  only  because  one  is  inclined  to  see  these 
as  immaterial  abstractions  hovering  over  the  terms  of  the  sentence. 
By  complementing  each  other,  the  two  principles  bear  out  my 
statements  relative  to  the  delimiting  of  units  (see  p.  103). 

Diachronic  Linguistics 

Chapter  I 

What  diachronic  linguistics  studies  is  not  relations  between  co- 
existing terms  of  a  language-state  but  relations  between  successive 
terms  that  are  substituted  for  each  other  in  time. 

There  is  really  no  such  thing  as  absolute  immobility  (see  pp. 
75  ff.).  Every  part  of  language  is  subjected  to  change.  To  each 
period  there  corresponds  some  appreciable  evolution.  Evolution 
may  vary  in  rapidity  and  intensity,  but  this  does  not  invahdate  the 
principle.  The  stream  of  language  flows  without  interruption; 
whether  its  course  is  calm  or  torrential  is  of  secondary  importance. 

That  we  often  fail  to  see  this  uninterrupted  evolution  is  due  to 
the  attention  paid  to  the  literary  language  which,  as  will  appear 
later  (pp.  195  ff.)  is  superimposed  on  the  vulgar  language  (i.e.  the 
natural  language)  and  is  subjected  to  other  forces.  The  literary 
language,  once  it  has  been  formed,  generally  remains  fairly  stable 
and  tends  to  keep  its  identity;  its  dependence  on  writing  gives  it 
special  guarantees  of  preservation;  therefore  it  cannot  show  us  how 
much  natural  languages  change  when  freed  from  any  literary 

Phonetics — and  all  of  phonetics — is  the  prime  object  of  dia- 
chronic linguistics.  In  fact,  the  evolution  of  sounds  is  incompatible 
with  the  notion  of  states ;  to  compare  phonemes  or  groups  of  pho- 
nemes with  what  they  were  previously  means  to  set  up  a  diachrony. 
One  period  may  be  closely  related  to  the  next,  but  when  the  two 
merge,  phonetics  ceases  to  play  a  part.  Nothing  is  left  but  the 
description  of  the  sounds  of  a  language-state,  and  that  is  the  task 
of  phonology. 

The  diachronic  character  of  phonetics  fits  in  very  well  with  the 



principle  that  anything  which  is  phonetic  is  neither  significant  nor 
grammatical  in  the  broad  sense  of  the  word  phonetic  (see  p.  18).  In 
studying  the  history  of  the  sounds  of  a  word,  we  may  ignore 
meaning  and,  by  considering  only  the  material  envelope  of  a  word, 
cut  out  phonic  slices  without  asking  whether  they  have  a  signi- 
fication. For  instance,  we  may  try  to  trace  the  meaningless  group 
-ewo-  in  Attic  Greek.  If  the  evolution  of  language  meant  nothing 
more  than  the  evolution  of  its  sounds,  the  opposition  between  the 
objects  that  belong  to  each  of  the  two  parts  of  linguistics  would 
immediately  be  crystal  clear.  It  would  be  obvious  that  diachronic 
is  equivalent  to  nongrammatical  and  synchronic  to  grammatical. 

But  sounds  are  not  the  only  things  that  change  with  time.  Words 
change  their  signification.  Grammatical  classes  evolve.  Some  of 
them  disappear  along  with  the  forms  that  were  used  to  express 
them  (e.g.  the  dual  number  in  Latin).  And  if  all  associative  and 
syntagmatic  facts  in  a  synchronic  state  have  their  history,  how 
is  the  absolute  distinction  between  diachrony  and  synchrony  to 
be  maintained?  This  becomes  very  difficult  when  we  leave  the 
domain  of  phonetics. 

It  is  worth  noting,  however,  that  many  changes  often  considered 
grammatical  are  really  only  phonetic.  Such  "grammatical"  cre- 
ations as  German  Hand:  Hdnde,  which  replaced  hant:  hanti  (see 
p.  83),  yield  completely  to  a  phonetic  explanation.  Another  pho- 
netic fact  is  at  the  base  of  compounds  of  the  type  Springbrun- 
nen,  Reitschule,  etc.  In  Old  High  German  the  first  element  was  not 
verbal  but  substantival.  Beta-hus  meant  'house  of  prayer';  but 
after  a  phonetic  change  brought  about  the  fall  of  the  final  vowel 
(beta  — >  bet-,  etc.),  a  semantic  contact  was  established  with  the 
verb  (beten,  etc.),  and  Bethaus  then  signified  'house  for  praying.' 

Something  similar  occurred  in  compounds  formed  with  the  word 
llch  'outward  appearance'  in  Old  High  German  (cf.  mannollch 
'having  the  appearance  of  a  man,'  redollch  'having  the  appearance 
of  reason,'  etc.).  Today,  in  a  number  of  adjectives  (cf.  verzeihlich, 
glaublich,  etc.),  -lich  is  comparable  to  the  suffix  in  pardon-able, 
heliev-able,  etc.,  and  at  the  same  time  the  interpretation  of  the 
first  element,  through  loss  of  the  final  vowel  (e.g.  redo  —^  red-),  is 
Ukened  to  a  verbal  root  (red-  from  reden) . 

In  glaublich,  glaub-  is  accordingly  linked  to  glauben  rather  than 


to  Glauhe,  and  in  spite  of  the  difference  in  the  radical,  sichtlich  is 
associated  with  sehen  and  not  Sicht. 

In  all  the  preceding  instances  and  in  many  other  similar  ones, 
the  distinction  between  the  two  classes  remains  clear-cut.  The  lin- 
guist must  keep  this  distinction  in  mind  or  risk  thinking  that  he  is 
studying  historical  grammar  when  he  is  actually  moving  succes- 
sively from  diachrony,  where  he  studies  phonetic  changes,  to 
synchrony,  where  he  examines  the  consequences  that  issue  from 
these  changes. 

But  this  restriction  does  not  remove  all  difficulties.  The  evolution 
of  any  grammatical  fact,  regardless  of  its  syntagmatic  or  gram- 
matical character,  is  not  like  the  evolution  of  a  sound.  It  is  not 
simple  but  decomposes  into  a  great  number  of  particular  facts  of 
which  only  a  part  are  phonetic.  In  the  genesis  of  a  syntagmatic  pat- 
tern like  the  French  future  prendre  ai  *(I)  have  to  take,'  which  be- 
came prendrai  *(I)  shall  take,'  there  are  at  least  two  distinct  facts, 
one  psychological  (the  synthesis  of  the  two  elements  of  the  concept) 
and  the  other  phonetic  and  dependent  on  the  first  (the  reduction 
of  the  two  accents  of  the  combination  to  one:  prendre 
ai  —>  prendrai). 

The  inflection  of  the  strong  Germanic  verb  (like  Modern  Ger- 
man gehen,  gab,  gegeben,  etc.,  cf.  Greek  leipo,  elipon,  leloipa,  etc.)  is 
based  chiefly  on  the  ablaut  of  radical  vowels.  These  alternations 
(see  p.  157),  which  began  as  a  relatively  simple  system,  doubtless 
result  from  a  mere  phonetic  fact.  But  for  the  oppositions  to  acquire 
such  functional  importance,  the  original  inflectional  system  had  to 
be  simplified  through  a  series  of  diverse  processes:  the  disappear- 
ance of  multiple  varieties  of  the  present  and  of  the  shades  of  mean- 
ing attached  to  them;  the  disappearance  of  the  imperfect,  the 
future,  and  the  aorist;  the  elimination  of  reduplication  of  the  per- 
fect, etc.  These  nonphonetic  changes  reduced  verbal  inflection  to  a 
restricted  group  of  forms  in  which  radical  alternations  became  very 
important  in  signaling  meaning.  Thus  the  opposition  e:  a  is  more 
significant  in  gehen:  gab  than  is  the  opposition  e:  o  in  Greek  leipo: 
leloipa,  for  the  German  perfect  lacks  reduplication  and  the  Greek 
has  it. 

Phonetic  change,  though  it  does  generally  affect  evolution  in 
some  way,  cannot  explain  it  entirely.  Once  the  phonetic  force  is 


eliminated,  we  find  a  residue  that  seems  to  justify  the  idea  of  a 
"history  of  grammar,"  and  therein  lies  the  real  difficulty.  This 
indispensable  distinction  between  diachrony  and  synchrony  would 
call  for  detailed  explanations  that  are  outside  the  scope  of  this 

In  the  following  chapters  we  shall  study,  successively,  phonetic 
changes,  alternation,  and  analogical  facts,  and  conclude  with  some 
remarks  about  folk  etymology  and  agglutination. 

Chapter  II 

1.  Their  Absolute  Regularity 

We  saw  earlier  (p.  93)  that  a  phonetic  change  affects  not  words 
but  sounds.  What  is  transformed  is  a  phoneme.  This  event,  though 
isolated  like  all  other  diachronic  events,  results  in  the  identical 
alteration  of  all  words  containing  the  same  phoneme.  It  is  in  this 
sense  that  phonetic  changes  are  absolutely  regular. 

In  German,  every  I  became  ei,  then  ai:  win,  trlben,  lihen,  zlt 
became  Wein,  treiben,  leihen,  Zeit;  every  u  became  au:  hus,  zun, 
ruch  became  Haus,  Zaun,  Rauch;  in  the  same  way  ii  changed  to  eu: 
hiiser  became  Hduser,  etc.  On  the  contrary,  the  diphthong  ie  be- 
came I,  which  is  still  written  ie:  cf.  biegen,  lieb,  Tier.  In  addition, 
every  uo  became  u:  muot  became  Mut,  etc.  Every  z  became  s  (writ- 

1  To  this  didactic  external  reason  might  be  added  another:  in  his  lectures 
F.  de  Saussure  never  approached  Unguistics  of  speaking  (see  pp.  17  ff.).  We 
recall  that  a  new  speech  form  always  owes  its  origin  to  a  series  of  individual 
facts  (see  p.  98).  We  might  say  that  the  author  refused  to  classify  these  as 
grammatical  in  the  sense  that  an  isolated  act  is  necessarily  foreign  to  language 
and  to  its  system,  which  depends  only  on  the  set  of  collective  patterns.  It  is 
only  when  an  innovation  becomes  engraved  in  the  memory  throuj^h  frequent 
repetition  and  enters  the  system  that  it  effects  a  shift  in  the  ocjuilibrium  of 
values  and  that  language  changes,  spontaneously  and  ipso  facto.  We  might 
apply  to  grammatical  evolution  what  was  said  on  pages  18  and  84  about 
phonetic  evolution:  its  end  result  is  outside  the  system,  for  the  system  is  never 
observed  in  its  evolution;  it  differs  from  one  moment  to  the  next.  This  at- 
tempted explanation  is  just  a  simple  suggestion  on  our  part.  [Ed.] 


ten  ss,  see  p.  36) :  wazer  — »  Wasser,  jliezen  -^  fliessen,  etc.  Every 
intervocalic  h  disappeared:  llhen,  sehen  — >  leien,  seen  (written 
leihen,  sehen).  Every  w  was  changed  to  labiodental  v  (written  w): 
wazer  -^  waser  (Wasser). 

In  French,  every  palatalized  I  became  y:  piller  'pillar'  and 
bouillir  'boil'  are  pronounced  ptTje,  huyir,  etc. 

In  Latin,  what  was  once  intervocalic  s  appears  as  r  in  another 
period:  *genesis,  *asena  -^  generis,  arena,  etc. 

Any  phonetic  change  at  all,  when  seen  in  its  true  light,  would 
confirm  the  perfect  regularity  of  these  transformations. 

2.  Conditioned  Phonetic  Changes 

The  preceding  examples  have  already  shown  that  phonetic  phe- 
nomena, far  from  always  being  absolute,  are  more  often  linked  to 
fixed  conditions.  Putting  it  another  way,  what  is  transformed  is 
not  the  phonological  species  but  the  phoneme  as  it  occurs  under 
certain  conditions — its  environment,  accentuation,  etc.  For  in- 
stance, s  became  r  in  Latin  only  between  vowels  and  in  certain 
other  positions;  elsewhere  it  remains  (cf.  est,  senex,  equos). 

Absolute  changes  are  extremely  rare.  That  changes  often  appear 
to  be  absolute  is  due  to  the  obscure  or  extremely  general  nature  of 
the  conditions.  In  German,  for  example,  i  became  ei,  ai,  but  only 
in  a  tonic  syllable.  Proto-Indo-European  A;i  became  h  in  Germanic 
(cf.  Proto-Indo-European  *k\olsom,  Latin  collum,  German  Hals), 
but  the  change  did  not  occur  after  s  (cf.  Greek  skotos  and  Gothic 
skadus  'shadow'). 

Besides,  the  classing  of  changes  as  absolute  or  conditioned  is 
based  on  a  superficial  view  of  things.  It  is  more  logical,  in  line  with 
the  growing  trend,  to  speak  of  spontaneous  and  combinatory  pho- 
netic phenomena.  Changes  are  spontaneous  when  their  cause  is 
internal  and  combinatory  when  they  result  from  the  presence  of 
one  or  more  other  phonemes.  The  passing  of  Proto-Indo-European 
0  to  Germanic  a  (cf.  Gothic  skadus,  German  Hals,  etc.)  is  thus  a 
spontaneous  fact.  Germanic  consonantal  mutations  or  Lautver- 
schiehungen  typify  spontaneous  change:  Proto-Indo-European  ki 
became  h  in  Proto-Germanic  (cf .  Latin  collum  and  Gothic  hals)  and 
Proto-Germanic  t,  which  is  preserved  in  English,  became  z  (pro- 
nounced ts)  in  High  German   (cf.  Gothic  taihun,  EngUsh  ten, 


German  zehn) .  Against  this,  the  passing  of  Latin  ct,  pi  to  Italian  tt 
(cf.  factum  -^fatto,  captlvum  — ^  cattivo)  is  a  combinatory  fact,  for 
the  first  element  was  assimilated  to  the  second.  The  German 
umlaut  is  also  due  to  an  external  cause,  the  presence  of  i  in  the 
following  syllable:  while  gast  did  not  change,  gasti  became  gesti, 

The  result  is  not  an  issue  in  either  case,  and  whether  or  not  there 
is  a  change  is  of  no  importance.  For  instance,  on  comparing  Gothic 
fisks  with  Latin  piscis  and  Gothic  skadus  with  Greek  skotos,  we 
observe  in  the  first  pair  the  persistence  of  i  and  in  the  second  the 
passing  of  o  to  a.  The  first  phoneme  remained  while  the  second  one 
changed,  but  what  matters  is  that  each  acted  independently. 

A  combinatory  phonetic  fact  is  always  conditioned,  but  a  spon- 
taneous fact  is  not  necessarily  absolute,  for  it  may  be  conditioned 
negatively  by  the  absence  of  certain  forces  of  change.  In  this  way 
Proto-Indo-European  ki  spontaneously  became  qu  in  Latin  (cf. 
quattuor,  inquillna,  etc.)  but  not,  for  instance,  when  followed  by 
0  or  M  (cf.  cottidie,  cold  secundus,  etc.).  In  the  same  way  the  per- 
sistence of  Proto-Indo-European  i  in  Gothic  fisks,  etc.  is  linked  to 
a  condition — the  i  could  not  be  followed  by  r  or  h,  for  then  it  be- 
came e,  written  at  (cf .  wair  -^  Latin  vir  and  maihsius  — >  German 

3.  Points  on  Method 

In  devising  formulas  to  express  phonetic  changes  we  must  con- 
sider the  preceding  distinctions  or  risk  presenting  the  facts 

Here  are  some  examples  of  inaccuracies. 

According  to  the  old  formulation  of  Verner's  law,  "in  Germanic 
every  noninitial  Ip  changed  to  6  if  the  accent  came  after  it":  cf.  on 
the  one  hand  *fa])er  — >  *fa'6er  (German  Vater),  *li]>ume  — »  *li'6ume 
(German  litten),  and  on  the  other  *^ris  (German  drei),  *bro])er 
(German  Bruder),  *li\>o  (German  hide),  where  J?  remains.  This 
formula  gives  the  active  role  to  accent  and  introduces  a  restrictive 
clause  for  initial  ]?.  What  actually  happened  is  quite  different.  In 
Germanic,  as  in  Latin,  J?  tended  to  sonorize  spontaneously  within 
a  word ;  only  the  placing  of  the  accent  on  the  preceding  vowel  could 
prevent  it.  Everything  is  therefore  reversed.  The  fact  is  spon- 


taneous,  not  combinatory,  and  the  accent  is  an  obstacle  rather  than 
the  precipitating  cause.  We  should  say:  "Every  internal  Ip  became 
tS  unless  the  change  was  opposed  by  the  placing  of  the  accent  on 
the  preceding  vowel." 

In  order  to  separate  what  is  spontaneous  from  what  is  com- 
binatory, we  must  analyze  the  stages  of  the  transformation  and 
not  mistake  the  mediate  result  for  the  immediate  one.  It  is  wrong 
to  explain  rhotacization,  for  instance  (cf.  Latin  *genesis  — >  generis), 
by  saying  that  s  became  r  between  two  vowels,  for  s,  having  no 
laryngeal  sound,  could  never  become  r  directly.  There  are  really 
two  acts.  First,  s  became  z  through  a  combinatory  change.  Second, 
this  sound  was  replaced  by  closely  related  r  since  z  had  not  been 
retained  in  the  sound  system  of  Latin.  The  second  change  was 
spontaneous.  It  is  therefore  a  serious  mistake  to  consider  the  two 
dissimilar  facts  as  a  single  phenomenon.  The  fault  is  on  the  one 
hand  in  mistaking  the  mediate  result  for  the  immediate  one  (s  — >  r 
instead  of  z-^r)  and  on  the  other,  in  regarding  the  total  phe- 
nomenon as  combinatory  when  this  is  true  of  only  its  first  part. 
This  is  the  same  as  saying  that  e  became  a  before  a  nasal  in  French. 
The  fact  is  that  there  were  in  succession  a  combinatory  change — 
nasalization  of  e  by  n  (cf .  Latin  ventum  — >  French  vent,  Latin 
femina  — »  French /ewa,  femd) — and  a  spontaneous  change  of  e  to  o 
(cf.  vant,  fdmd,  now  vd,  fdm).  To  raise  the  objection  that  the  change 
could  occur  only  before  a  nasal  consonant  would  be  pointless.  The 
question  is  not  why  e  was  nasalized  but  only  whether  the  trans- 
formation of  e  into  d  is  spontaneous  or  combinatory. 

The  most  serious  mistake  in  method  that  I  can  recall  at  this 
point — although  it  is  not  related  to  the  principles  stated  above — 
is  that  of  formulating  a  phonetic  law  in  the  present  tense,  as  if  the 
facts  embraced  by  it  existed  once  and  for  all  instead  of  being  born 
and  dying  within  a  span  of  time.  The  result  is  chaos,  for  in  this  way 
any  chronological  succession  of  events  is  lost  sight  of.  I  have  al- 
ready emphasized  this  point  (p.  97)  in  analyzing  the  successive 
phenomena  that  explain  the  duality  of  trikhos:  thriksi.  Whoever 
says  "s  became  r  in  Latin"  gives  the  impression  that  rhotacization 
is  inherent  in  the  nature  of  language  and  finds  it  difficult  to  account 
for  exceptions  like  causa,  rlsus,  etc.  Only  the  formula  "intervocalic 
s  became  r  in  Latin"  justifies  our  believing  that  causa,  rlsus,  etc. 


had  no  s  at  the  moment  when  s  became  r  and  were  sheltered  from 
change.  The  fact  is  that  speakers  still  said  caussa,  rlssus,  etc.  For 
a  similar  reason  we  must  say  "a  became  e  in  the  Ionian  dialect  (cf. 
mater  meter,  etc.),  for  otherwise  we  would  not  know  what  to  make 
of  forms  like  pdsa,  phdsi,  etc.  (which  were  still  pansa,  phansi,  etc. 
during  the  period  of  the  change), 

4.  Causes  of  Phonetic  Changes 

The  search  for  the  causes  of  phonetic  changes  is  one  of  the  most 
difficult  problems  of  linguistics.  Many  explanations  have  been 
proposed,  but  none  of  them  thoroughly  illuminates  the  problem. 

1)  One  supposition  is  that  racial  predispositions  trace  before- 
hand the  direction  of  phonetic  changes.  This  raises  a  question  of 
comparative  anthropology:  Does  the  phonational  apparatus  vary 
from  one  race  to  the  next?  No,  scarcely  more  than  from  one  in- 
dividual to  the  next,  A  newborn  Negro  transplanted  to  France 
speaks  French  as  well  as  a  native  Frenchman.  Furthermore,  ex- 
pressions like  "the  Italian  vocal  apparatus"  or  "the  mouth  of 
Germanic  speakers  does  not  allow  that"  imply  that  a  mere  histori- 
cal fact  is  a  permanent  characteristic.  This  is  similar  to  the  mistake 
of  stating  a  phonetic  law  in  the  present  tense.  To  pretend  that  the 
Ionian  vocal  apparatus  finds  long  a  difficult  and  changes  it  to  e  is 
just  as  erroneous  as  to  say  that  d  "becomes"  e  in  Ionian. 

The  Ionian  vocal  apparatus  had  no  aversion  to  d,  for  this  sound 
was  used  in  certain  instances.  This  is  obviously  an  example,  not 
of  racial  incapacity,  but  of  a  change  in  articulatory  habits.  In  the 
same  way  Latin,  which  had  not  retained  intervocalic  s  {*genesis  — >• 
generis),  reintroduced  it  a  short  time  later  (cf.  *rissus  — >  risus). 
These  changes  do  not  indicate  a  permanent  disposition  of  the 
Latin  voice. 

There  is  doubtless  a  general  direction  that  phonetic  phenomena 
follow  during  a  particular  period  and  within  a  specific  nation.  The 
monophthongizations  of  diphthongs  in  Modern  French  are  mani- 
festations of  one  and  the  same  tendency,  but  we  would  find  similar 
general  currents  in  political  history  and  never  question  their  being 
merely  historical  without  any  direct  influence  of  race, 

2)  Phonetic  changes  have  often  been  considered  adaptions  to 
conditions   of   soil   and   climate.    Consonants   abound   in   some 


northern  languages  while  more  vowels  occur  in  certain  southern 
languages,  giving  them  their  harmonious  sound.  Climate  and  living 
conditions  may  well  influence  language,  but  the  problem  becomes 
complicated  as  soon  as  we  enter  into  detail:  beside  the  Scandi- 
navian idioms  with  their  many  consonants  are  those  of  the  Lapps 
and  Finns,  which  are  even  more  vocalic  than  Italian.  We  also 
notice  that  the  accumulation  of  consonants  in  present-day  German 
is  in  many  instances  a  quite  recent  fact,  due  to  the  fall  of  posttonic 
vowels;  that  certain  dialects  of  southern  France  are  less  opposed 
to  consonantal  clusters  than  the  French  of  the  north ;  that  Serbian 
has  as  many  consonantal  clusters  as  Great  Russian,  etc. 

3)  The  cause  of  phonetic  changes  has  also  been  ascribed  to  the 
law  of  least  effort  by  which  two  articulations  are  replaced  by  one 
or  a  difficult  articulation  by  an  easier  one.  This  idea,  regardless  of 
what  is  said  about  it,  is  w^orth  examining.  It  may  clarify  the  cause 
of  phonetic  changes  or  at  least  indicate  the  direction  that  the 
search  for  it  must  take. 

The  law  of  least  effort  seems  to  explain  a  certain  number  of  cases : 
the  passing  of  an  occlusive  to  a  spirant  (Latin  habere  — >  French 
avoir  'have') ;  the  fall  of  great  clusters  of  final  syllables  in  many 
languages;  phenomena  relating  to  assimilation  (e.g.  ly  — >  II  as  in 
*alyos  — ^  Greek  alios,  tn  — >  nn  as  in  *atnos  — >  Latin  annus) ;  the 
monophthongization  of  diphthongs,  which  is  only  another  type  of 
assimilation  (e.g.  ai  —^  e  as  in  French  maizon  — >  mezo,  written 
maison  'house'),  etc. 

But  we  might  mention  just  as  many  instances  where  exactly  the 
opposite  occurs.  Against  monophthongization,  for  example,  we  can 
set  the  change  of  German  l,  u,  ii,  to  ei,  au,  eu.  If  the  shortening  of 
Slavic  a,  etod,  e  is  due  to  least  effort,  then  the  reverse  phenomenon 
offered  by  German  (fater  -^  Vdter,  gehen  — »  geben)  must  be  due  to 
greatest  effort.  If  voicing  is  easier  than  nonvoicing  (cf .  opera  — > 
Provencal  obra),  the  reverse  must  necessitate  greater  effort,  and  yet 
Spanish  passed  from  z  to  X  (cf .  hixo,  written  hijo)  and  Germanic 
changed  b,  d,  gtop,t,k.  If  loss  of  aspiration  (cf.  Proto-Indo-Euro- 
pean *bherd  — >  Germanic  beran)  is  considered  a  lessening  of  effort, 
what  is  to  be  said  of  German,  which  inserts  aspiration  where  it  did 
not  exist  {Tanne,  Pute,  etc.,  pronounced  Thanne,  Phute)? 

The  foregoing  remarks  do  not  pretend  to  refute  the  proposed 


solution.  In  fact,  we  can  scarcely  determine  what  is  easiest  or  most 
difficult  for  each  language  to  pronounce.  Shortening  means  less 
effort  in  the  sense  of  duration,  but  it  is  equally  true  that  long 
sounds  allow  careless  pronunciations  while  short  sounds  require 
more  care.  Given  different  predispositions,  we  can  therefore  pre- 
sent two  opposing  facts  from  the  same  viewpoint.  Thus  where  k 
became  ts  (cf.  Latin  cedere  — >  Italian  cedere),  there  is  apparently  an 
increase  in  effort  if  we  consider  only  the  end  terms  of  the  change, 
but  the  impression  would  probably  differ  if  we  reconstructed  the 
chain:  k  became  palatalized  k'  through  assimilation  to  the  folloA\'ing 
vowel ;  then  k'  passed  to  ky;  the  pronunciation  did  not  become  more 
difficult;  two  tangled  elements  in  k'  were  clearly  differentiated; 
then  from  ky  speakers  passed  successively  to  ty,  tx,  t^,  everywhere 
with  less  effort. 

The  law  of  least  effort  would  require  extensive  study.  It  would 
be  necessary  to  consider  simultaneously  the  physiological  view- 
point (the  question  of  articulation)  and  the  psychological  view- 
point (the  question  of  attention). 

4)  An  explanation  that  has  been  favored  for  several  years 
attributes  changes  in  pronunciation  to  our  phonetic  education 
during  childhood.  After  much  groping  and  many  trials  and  cor- 
rections, the  child  succeeds  in  pronouncing  what  he  hears  around 
him;  here  would  be  the  starting  point  of  all  changes;  certain  un- 
corrected inaccuracies  would  win  out  in  the  individual  and  become 
fixed  in  the  generation  that  is  growing  up.  Children  often  pro- 
nounce t  for  k,  and  our  languages  offer  no  corresponding  phonetic 
change  in  their  history.  But  this  is  not  true  of  other  deformations. 
In  Paris,  for  instance,  many  children  pronounce  fl'eur  (fleur 
'flower')  and  Wane  (blanc  'white')  with  palatalized  I;  now  it  was 
through  a  similar  process  that  florem  became  ft'ore,  then  fiore,  in 

The  preceding  observations  deserve  careful  attention  but  leave 
the  problem  undented.  Indeed,  what  prompts  a  generation  to 
retain  certain  mistakes  to  the  exclusion  of  others  that  are  just  as 
natural  is  not  clear.  From  all  appearances  the  choice  of  faulty  pro- 
nunciations is  completely  arbitrary,  and  there  is  no  obvious  reason 
for  it.  Besides,  why  did  the  phenomenon  break  through  at  one  time 
rather  than  another? 


The  same  question  applies  to  all  the  preceding  causes  of  phonetic 
changes  if  they  are  accepted  as  real.  Climatic  influence,  racial  pre- 
disposition, and  the  tendency  toward  least  effort  are  all  permanent 
or  lasting.  Why  do  they  act  sporadically,  sometimes  on  one  point 
of  the  phonological  system  and  sometimes  on  another?  A  historical 
event  must  have  a  determining  cause,  yet  we  are  not  told  what 
chances  in  each  instance  to  unleash  a  change  whose  general  cause 
has  existed  for  a  long  time.  This  is  the  most  diflEicult  point  to 

5)  Phonetic  changes  are  sometimes  linked  to  the  general  state 
of  the  nation  at  a  particular  moment.  Languages  go  through  some 
periods  that  are  more  turbulent  than  others.  There  have  been 
attempts  to  relate  phonetic  changes  to  turbulent  periods  in  a 
nation's  history  and  in  this  way  to  discover  a  link  between  political 
instability  and  linguistic  instability;  this  done,  some  think  that 
they  can  apply  conclusions  concerning  language  in  general  to 
phonetic  changes.  They  observe,  for  example,  that  the  sharpest 
upheavals  of  Latin  in  its  development  into  the  Romance  languages 
coincided  with  the  highly  disturbed  period  of  invasions.  Two  dis- 
tinctions will  serve  as  guideposts: 

a)  Political  stability  does  not  influence  language  in  the  same  way 
as  political  instability ;  here  there  is  no  reciprocity.  When  poUtical 
equilibrium  slows  down  the  evolution  of  language,  a  positive 
though  external  cause  is  involved.  But  instability,  which  has  the 
opposite  effect,  acts  only  negatively.  ImmobiUty — the  relative 
fixation  of  an  idiom — may  have  an  external  cause  (the  influence 
of  a  court,  school,  an  academy,  writing,  etc.)  which  in  turn  is  posi- 
tively favored  by  social  and  political  equilibrium.  But  if  some 
external  upheaval  that  has  affected  the  equihbrium  of  the  nation 
precipitates  Hnguistic  evolution,  this  is  because  language  simply 
reverts  back  to  its  free  state  and  follows  its  regular  course.  The 
immobility  of  Latin  of  the  classical  period  is  due  to  external  facts; 
the  changes  that  it  later  underwent,  however,  were  self-generated 
in  the  absence  of  certain  external  conditions. 

b)  Here  we  are  dealing  only  with  phonetic  phenomena  and  not 
with  every  type  of  modification  of  language.  Grammatical  changes 
are  obviously  similar.  Because  they  are  always  closely  linked  to 


thought,  grammatical  facts  are  more  easily  affected  by  the  impact 
of  external  upheavals,  which  have  a  more  immediate  repercussion 
on  the  mind.  But  there  is  no  solid  basis  for  the  behef  that  sudden 
evolutions  of  the  sounds  of  an  idiom  correspond  to  turbulent 
periods  in  the  history  of  a  nation. 

Still,  it  is  impossible  to  cite  a  single  period — even  among  those 
where  language  is  in  a  deceptive  state  of  immobility — that  has 
witnessed  no  phonetic  changes. 

6)  The  "linguistic  substratum"  has  also  been  posited  as  the 
cause  of  phonetic  changes.  The  absorption  of  an  indigenous  popu- 
lation by  newcomers  brings  about  certain  changes.  The  difference 
between  Proven9al  and  French  {langue  d'oc  and  langue  dfoil)  would 
accordingly  correspond  to  a  different  proportion  of  the  autoch- 
thonous Celtic  element  in  the  two  parts  of  Gaul.  This  theory  has 
also  been  used  to  trace  the  dialectal  differences  of  Italian  and  the 
influence  of  Ligurian,  Etruscan,  etc.,  depending  on  the  region.  But 
first,  this  hypothesis  supposes  circumstances  that  are  rarely  found. 
Second,  one  must  be  more  specific :  Did  earlier  populations  intro- 
duce some  of  their  own  articulatory  habits  into  the  new  language 
on  adopting  it?  This  is  admissible  and  quite  natural.  But  if  the 
imponderable  forces  of  race,  etc.  are  called  in  anew,  the  pitfalls 
described  earlier  reappear. 

7)  A  final  explanation — which  scarcely  merits  the  name — com- 
pares phonetic  changes  to  changes  in  fashion.  But  no  one  has 
explained  these  changes.  We  know  only  that  they  depend  on  laws 
of  imitation,  which  are  the  concern  of  the  psychologist.  This  ex- 
planation, though  it  does  not  solve  our  problem,  has  the  advantage 
of  fitting  it  into  another  larger  problem  and  positing  a  psychologi- 
cal basis  for  phonetic  changes.  But  where  is  the  starting  point  of 
imitation?  That  is  the  mystery,  in  phonetic  changes  as  well  as  in 
changes  of  fashion. 

5.  The  Effect  of  Phonetic  Changes  Is  Unlimited 

If  we  try  to  determine  how  far  phonetic  changes  will  go,  we  see 
immediately  that  they  are  unhmited  and  incalculable,  i.e.  we  can- 
not foresee  where  they  will  stop.  It  is  childish  to  think  that  the 
word  can  be  changed  only  up  to  a  certain  point,  as  if  there  were 
something  about  it  that  could  preserve  it.  Phonetic  modifications 


derive  their  character  from  the  arbitrariness  of  the  linguistic  sign,^ 
which  is  distinct  from  the  signified. 

We  can  easily  observe  that  the  sounds  of  a  word  have  been 
affected  at  a  certain  moment  and  see  the  extent  of  the  damage,  but 
we  cannot  say  beforehand  how  far  the  word  has  become  or  will 
become  unrecognizable. 

Like  every  word  having  the  same  ending,  Proto-Indo-European 
*aiwom  (cf .  Latin  aevom)  changed  to  *aiwan,  *aiwa,  *aiw  in  Proto- 
Germanic;  next,  *aiw  became  ew  in  Old  High  German,  as  did  every 
word  that  contained  the  cluster  aiw;  then  the  change  of  final  wtoo 
resulted  in  eo,  which  in  turn  passed  to  eo,  io  in  accordance  with 
other  equally  general  rules;  finally  io  became  ie,  je,  giving  Modern 
German  je  (cf.  das  schonste,  was  ichje  gesehen  habe  'the  prettiest 
that  I  have  ever  seen'). 

The  modern  word  does  not  contain  a  single  one  of  its  original 
elements  when  considered  from  the  viewpoint  of  the  starting  point 
and  the  end  result.  Each  step,  when  viewed  separately,  is  abso- 
lutely certain  and  regular  and  limited  in  its  effect;  viewed  as  a 
whole,  however,  the  word  gives  the  impression  of  an  unlimited 
number  of  modifications.  We  might  make  the  same  observation 
about  Latin  calidum  by  first  leaving  out  the  transitional  forms 
and  comparing  this  form  with  Modern  French  so  (written  chaud 
'warm'),  then  retracing  the  steps:  calidum,  calidu,  caldu,  cold,  colt, 
tsalt,  tsaut,  Saut,  ^ot,  ^o.  Compare  also  Vulgar  Latin  *waidanju 
— >  ge  (written  gain  'gain'),  minus  —^  mwe  (written  moins  'less'), 
hoc  nil  -^  wi  (written  oui  'yes'). 

A  phonetic  change  is  also  unlimited  and  incalculable  in  that  it 
affects  all  types  of  signs,  making  no  distinction  between  radicals, 
suffixes,  etc.  This  must  be  true  a  priori,  for  if  grammar  interfered, 
the  phonetic  phenomenon  would  mingle  with  the  synchronic  fact, 
a  thing  that  is  radically  impossible.  It  is  in  this  sense  that  we  can 
speak  of  the  blind  nature  of  the  evolutions  of  sounds. 

For  instance,  s  fell  in  Greek  after  n  not  only  in  *khdnses  'geese,' 
*menses  'months'  (giving  khenes,  mtnes),  where  it  had  no  gram- 
matical value,  but  also  in  verbal  forms  like  *etensa,  *ephansa,  etc. 
(giving  eteina,  ephena,  etc.),  where  it  marked  the  aorist.  In  Middle 
High  German  the  posttonic  vowels  i,  e,  a,  o  regularly  became  e 

"  Meaning  signifier.  See  p.  75,  note.  [Tr.] 


(gihil  — ^  Giebel,  meistar  — >  Meister)  even  though  the  difference  in 
timbre  marked  a  number  of  inflectional  endings;  that  is  how  the 
accusative  singular  hoion  and  the  genitive  and  dative  singular  hoten 
merged  into  hoten. 

Phonetic  changes  will  thus  cause  a  profound  disturbance  in  the 
grammatical  organism  if  they  are  not  stopped  by  some  barrier. 
This  will  be  the  subject  matter  of  the  next  chapter. 

Chapter  III 


1.  The  Breaking  of  the  Grammatical  Bond 

One  of  the  first  consequences  of  the  phonetic  phenomenon  is  the 
breaking  of  the  grammatical  bond  that  unites  two  or  more  terms. 
The  result  is  that  one  word  is  no  longer  felt  to  be  derived  from 
another : 

mansio — *mansidndticus 

maison  'house'  ||  menage  'housekeeping' 

The  collective  mind  of  the  community  of  speakers  formerly  saw 
*mansid-ndticus  as  a  derivative  of  mansio;  then  phonetic  vicissi- 
tudes separated  them.  Similarly: 

(vervex — vervecdrius) 

Vulgar  Latin  berblx — herblcdrius 

brebis  'ewe'  ||  berger  'shepherd' 

The  separation  naturally  has  its  countereffect  on  value.  In 
certain  local  dialects  berger  means  specifically  'a  herder  of  oxen.' 
Other  examples : 

Grdtidnopolis—grdtidnopolitdnus  \\\\\  decem — undecim 
Grenoble  \\  Gresivaudan  |||{|||||||||||  dix  'ten'  ||  onze  'eleven' 


Gothic  bitan  'bite' — hitum  'we  have  bitten' — hitr  'bitter,  biting'  is 
a  similar  example.  Following  the  change  of  t  to  ts  (2)  on  the  one 
hand  and  the  conservation  of  the  cluster  tr  on  the  other,  West 
Germanic  had  blzan,  bizum  \  \  hitr. 

In  addition,  phonetic  evolution  may  break  the  normal  relation 
between  two  inflected  forms  of  the  same  word.  In  Old  French,  for 
instance,  comes — comiiem  became  cuens  \  \  comte,  bard — baronem  — > 
ber  II  baron,  presbiter — presbiterum  — >  prestre  \\  provoire. 

Or  an  ending  may  split  in  two.  All  accusative  singulars  were 
characterized  by  the  same  final  -m  in  Proto-Indo-European 
(*eki  worn,  *owim,  *podm,  *mdter'm,  etc.).'  In  Latin  there  was  no 
radical  change  in  this  respect,  but  in  Greek  the  very  different  treat- 
ment of  the  sonant  and  con-sonant  nasal  created  two  distinct 
series  of  iorms'.hippon,  6{w)in  against  poda,  matera.  The  accusative 
plural  evinces  a  similar  fact  (cf.  hippous  and  podas). 

2.  Effacement  of  the  Structure  of  Words 

Another  grammatical  effect  of  phonetic  changes  is  that  the  dis- 
tinct parts  that  helped  to  fix  the  value  of  a  word  become  un- 
analyzable.  The  word  becomes  an  indivisible  whole.  Examples: 
French  ennemi  'enemy'  (cf.  Latin  in-imlcus — amicus);  Latin 
perdere  (cf.  older  per-dare — dare),  amicio  (for  *ambjacio — jacio); 
German  Drittel  (for  drit-teil — Teil) . 

Effacement  of  the  structure  of  words  is  obviously  related  at 
several  points  to  the  breaking  of  grammatical  bonds  (see  Section  1 
above).  For  instance,  stating  that  ennemi  cannot  be  analyzed  is 
another  way  of  saying  that  its  parts  can  no  longer  be  compared  as 
in  in-imlcus  from  simple  amicus.  The  formula : 

amicus — inimicus 
ami  II  ennemi 

is  very  similar  to : 

mansio — mansiondticus 
maison  \\  menage. 

Cf .  also :  decem — undecim  against  dix  \  \  onze. 
3  Or  -n?  See  p.  92,  note.  [Ed.] 


The  simple  Classical  Latin  forms  hunc,  hanc,  hdc,  etc.  go  back 
to  hon-ce,  han-ce,  ha-ce,  etc.  (attested  by  epigraphic  forms)  and  are 
the  result  of  the  agglutination  of  a  pronoun  with  a  particle  -ce. 
Once  hon-ce,  etc.  could  be  compared  with  ec-ce,  etc.,  but  com- 
parison was  no  longer  possible  after  -e  had  fallen.  That  is  just 
another  way  of  saying  that  the  elements  of  hunc,  hanc,  hdc,  etc. 
are  no  longer  distinct. 

Phonetic  evolution  first  obscures  analysis,  then  makes  it  com- 
pletely impossible.  The  inflection  of  nouns  in  Proto-Indo-European 
is  a  case  in  point. 

The  Proto-Indo-European  declension  was  as  follows :  nominative 
singular  *pod-s,  accusative  *pod-m,  dative  *pod-ai,  locative  *pod-i, 
nominative  plural  *pod-es,  accusative  *pod-ns,  etc.  At  first  the 
inflection  of  *ek\Wos  was  identical:  *ekiWo-s,  *ekiWo-m,  *ekiWO-ai, 
*ekiWo-i,  *ekiWO-es,  *ekiWO-ns,  etc. ;  during  that  period  *€kiwo-  was 
singled  out  as  easily  as  *pod-.  But  vocalic  contractions  later  modi- 
fied that  state,  giving  dative  *ek\Woi,  locative  *ek\Woi,  nominative 
plural  *ekiWos.  From  that  moment  the  distinctness  of  the  radical 
*ekiWO-  was  compromised  and  its  analysis  became  elusive.  Still 
later,  new  changes  like  the  differentiation  between  accusatives 
(see  p.  154)  wiped  out  the  last  trace  of  the  original  state.  The  con- 
temporaries of  Xenophon  probably  had  the  impression  that  the 
radical  was  hipp-  and  that  the  inflectional  endings  were  vocahc 
(hipp-os,  etc.),  with  the  result  that  the  endings  of  words  Uke 
*ekiWO-s  and  *pod-s  were  distinct.  In  inflection  as  elsewhere,  any- 
thing which  interferes  with  analysis  helps  to  loosen  grammatical 

3.  There  Are  No  Phonetic  Doublets 

In  the  two  cases  that  we  have  examined  (Sections  1  and  2), 
evolution  radically  separated  two  terms  that  originally  were  united 
grammatically.  This  phenomenon  might  give  rise  to  a  serious  mis- 
take in  interpretation. 

On  observing  the  relative  identity  of  Vulgar  Latin  bard:  baronem 
and  the  dissimilarity  of  Old  French  ber:  baron,  is  one  not  justified 
in  saying  that  one  and  the  same  original  unit  (bar-)  developed  in 
divergent  directions  and  produced  two  forms?  No,  for  the  same  unit 
cannot  be  subjected  at  the  same  time  and  in  the  same  place  to  two 


different  transformations;  that  would  be  contrary  to  the  very  defi- 
nition of  phonetic  changes.  By  itself,  phonetic  evolution  cannot 
create  two  forms  to  replace  one. 

Here,  introduced  by  way  of  examples,  are  the  objections  that 
might  be  raised  against  my  thesis : 

Collocdre  gave  both  coucher  'sleep'  and  colloquer  'place,'  someone 
might  say.  No,  it  gave  only  coucher;  colloquer  is  only  a  learned 
borrowing  from  Latin  (cf.  rangon  'ransom'  and  redemption  're- 
demption') . 

Another  objection  might  be  that  cathedra  gave  two  authentic 
French  words,  chaire  'pulpit'  and  chaise  'chair.'  The  fact  that 
chaise  is  a  dialectal  form  is  forgotten.  The  Parisian  dialect  changed 
intervocalic  r  to  z.  For  instance,  speakers  said  pese,  mese  for  pere 
'father,'  mkre  'mother';  literary  French  has  kept  only  two  speci- 
mens of  the  localism:  chaise  and  besides,  the  doublet  of  hericles 
'spectacles,'  derived  from  heryl  'beryl.'  The  same  is  true  of  Picard 
r escape  'one  who  has  escaped  (death  or  injury),'  which  has  just 
gained  currency  in  French  and  now  stands  in  contrast  to  rechappe 
'one  who  has  (voluntarily)  escaped  (from  confinement).'  French 
cavalier  'rider'  and  chevalier  'knight'  and  cavalcade  'ride'  and 
chevauchee  'distance  traversed'  are  found  side  by  side  simply  be- 
cause cavalier  and  cavalcade  were  borrowed  from  Italian.  The 
development  of  calidum,  which  became  chaud  'warm'  in  French 
and  caldo  in  Italian,  is  essentially  the  same.  All  the  foregoing 
examples  are  instances  of  borrowings. 

The  answer  to  the  objection  that  the  Latin  pronoun  me  resulted 
in  two  forms  in  French,  me  and  moi  (cf .  il  me  voit  'he  sees  me'  and 
c'est  moi  qu'il  voit  'it's  me  that  he  sees')  is  this:  unstressed  Latin 
me  became  me  while  stressed  me  became  moi;  now  the  presence  or 
absence  of  stress  depends,  not  on  the  phonetic  laws  that  made  me 
become  me  and  moi,  but  on  the  function  of  the  word  in  the  sen- 
tence; it  is  a  grammatical  duality.  In  the  same  way,  German 
*ur-  remained  ur-  when  stressed  and  became  er-  when  protonic 
(cf .  iirlauh  and  erlauhen) ;  but  the  functioning  of  the  accent  is  itself 
linked  to  the  structural  patterns  that  contained  ur-  and  thus  to 
a  grammatical  and  synchronic  condition.  Finally,  to  come  back 
to  the  first  example,  differences  of  form  and  accent  in  the  pair 
hard:  haronem  evidently  antedate  phonetic  changes. 


In  fact,  phonetic  doublets  do  not  exist.  The  evolution  of  sounds 
only  emphasizes  previous  differences.  Wherever  these  differences 
are  not  due  to  external  causes  (as  in  borrowings),  they  imply  gram- 
matical and  synchronic  dualities  that  are  absolutely  unrelated  to 
phonetic  changes. 

4.  Alternation 

Two  words  like  maison:  menage  seldom  tempt  us  to  try  to  dis- 
cover what  is  responsible  for  the  difference,  either  because  the 
differential  elements  {-ezo  and  -en-  do  not  lend  themselves  well  to 
comparison,  or  because  no  other  pair  offers  a  parallel  opposition. 
But  often  it  happens  that  the  two  related  words  differ  in  only  one 
or  two  elements  which  are  easily  singled  out,  and  that  the  same 
difference  is  regularly  repeated  in  a  series  of  like  pairs;  this  is 
alternation,  the  largest  and  most  common  of  the  grammatical  facts 
in  which  phonetic  changes  play  a  part. 

In  French,  every  Latin  o  in  an  open  syllable  became  eu  when 
stressed  and  ou  when  protonic;  this  produced  pairs  like  pouvons 
'(we)  can':  peuvent  '(they)  can,'  oeuvre  'work':  ouvrier  'worker,' 
nouveau:  neuf  'new,'  etc.,  where  it  is  easy  to  single  out  a  differential 
and  regularly  variable  element.  In  Latin,  rhotacization  causes 
gero  to  alternate  with  gestus,  oneris  with  onus,  maeor  with  maestus, 
etc.  Since  s  was  treated  differently  according  to  the  position  of  the 
accent  in  Germanic,  Middle  High  German  has  ferliesen:  ferloren, 
kiessen:  gekoren,  friesen:  gefroren,  etc.  The  fall  of  Proto-Indo- 
European  e  is  reflected  in  Modern  German  in  the  oppositions 
heissen:  hiss,  leiden:  litt,  reiten:  ritt,  etc. 

In  all  the  preceding  examples  the  radical  element  is  the  part  that 
is  affected.  But  of  course  all  parts  of  a  word  may  have  similar 
oppositions.  Nothing  is  more  common,  for  instance,  than  a  prefix 
that  takes  different  forms  according  to  the  make-up  of  the  first  part 
of  the  radical  (cf.  Greek  apo-didomi:  ap-erchomai,  French  inconnu 
'unknowTi' :  inutile  'useless').  The  Proto-Indo-European  alternation 
e:  0,  which  certainly  must,  in  the  last  analysis,  have  a  phonetic 
basis,  is  found  in  a  great  number  of  suffixal  elements  (Greek  hippos: 
hippe,  pher-o-men:  pher-e-te,  gen-os:  gen-e-os  for  *gen-es-os,  etc.). 
Old  French  gives  special  treatment  to  Latin  accented  a  after 
palatals;  this  results  in  an  e:  ie  alternation  in  a  number  of  in- 


flectional  endings  (cf.  chant-er:  jug-ier,  chant-e;  jug-ie,  chant-ez: 
jug-iez,  etc.). 

Alternation  is  then  defined  as  a  correspondence  existing  between 
two  definite  sounds  or  groups  of  sounds  and  shifting  regularly  between 
two  series  of  coexisting  forms. 

Phonetic  changes  alone  do  not  explain  doublets,  and  are  ob- 
viously neither  the  sole  cause  nor  the  main  cause  of  alternation. 
Whoever  says  that  Latin  nov-  became  neuv-  and  nouv-  (French 
neuve  and  nouveau)  through  a  phonetic  change  is  fabricating  an  im- 
aginary unity  and  failing  to  see  a  pre-existing  synchronic  duality. 
The  different  position  of  nov-  in  nov-us  and  nov-ellus  is  both  ante- 
cedent to  the  phonetic  change  and  distinctly  grammatical  (cf. 
baro:  bar  mem).  The  synchronic  duality  is  what  originates  and 
makes  possible  any  alternation.  The  phonetic  phenomenon  broke 
no  unity;  it  merely  made  an  opposition  between  coexisting  terms 
more  obvious  by  discarding  certain  sounds.  It  is  a  mistake — and 
one  shared  by  many  linguists — to  assume  that  alternation  is  pho- 
netic simply  because  sounds  make  up  its  substance  and  play  a  part 
in  its  genesis  through  their  alterations.  The  fact  is  that  alternation, 
whether  considered  from  its  starting  point  or  end  result,  is  always 
both  grammatical  and  synchronic. 

5.  Laws  of  Alternation 

Can  alternation  be  reduced  to  laws?  If  so,  what  is  the  nature  of 

these  laws? 

Take  the  alternation  e:  i,  which  occurs  so  frequently  in  Modern 
German.  If  we  lump  all  examples  together  and  consider  them  in- 
discriminately (geben:  gibt,  Feld:  Gefilde,  Wetter:  wittern,  helfen: 
Hilfe,  sehen:  Sicht,  etc.),  we  can  formulate  no  general  principle. 
But  if  we  extract  from  this  mass  the  pair  geben:  gibt  and  set  it  in 
opposition  to  schelten:  schilt,  helfen:  hilft,  nehmen:  nimmt,  etc.,  we 
see  that  the  alternation  coincides  with  distinctions  of  tense,  person, 
etc.  In  lang:  Ldnge,  stark:  Starke,  hart:  Hdrte,  etc.,  a  similar  oppo- 
sition is  linked  to  the  formation  of  substantives  from  adjectives; 
in  Hand:  Hdnde,  Gast:  Gdste,  etc.,  to  the  formation  of  the  plural, 
and  so  on  for  all  the  many  cases  that  Germanic  students  class 
under  ablaut  (consider  also  finden:  fand,  or  finden:  Fund,  binden: 
band,  or  binden:  Bund,  schiessen:  schoss:  Schuss,  fliessen:  floss: 


Fluss,  etc.).  Ablaut,  or  radical  vocalic  variation  coinciding  with  a 
grammatical  opposition,  is  a  prime  example  of  alternation  but  is 
distinguished  from  the  general  phenomenon  by  no  particular 

Ordinarily,  then,  alternation  is  distributed  regularly  among 
several  terms  and  coincides  with  an  important  opposition  of 
function,  class,  or  determination.  It  is  possible  to  speak  of  gram- 
matical laws  of  alternation,  but  these  laws  are  only  a  fortuitous 
result  of  the  underlying  phonetic  facts.  When  phonetic  facts  create 
a  regular  opposition  between  two  series  of  terms  that  have  an  op- 
position of  value,  the  mind  seizes  upon  the  material  difference, 
gives  it  significance,  and  makes  it  the  carrier  of  the  conceptual 
difference  (see  pp.  84  ff.).  The  laws  of  alternation,  like  all  syn- 
chronic laws,  are  simple  structural  principles;  they  are  not  im- 
perative. It  is  completely  wrong  to  say,  as  people  so  readily  do, 
that  the  a  of  Nacht  changes  to  a  in  the  plural  Nachte,  for  this  gives 
the  illusion  that  a  transformation  governed  by  an  imperative 
principle  comes  between  one  term  and  the  next.  What  we  are  ac- 
tually dealing  with  is  a  simple  opposition  of  forms  resulting  from 
phonetic  evolution.  To  be  sure  analogy  (to  be  considered  later  in 
Chapter  VI)  may  create  new  pairs  that  show  the  same  phonic 
difference  (cf.  Kranz:  Krdnze,  modeled  on  Gast:  Gdste,  etc.).  The 
law  thus  seems  to  apply  like  a  rule  that  governs  usage  to  the  extent 
of  modifying  it.  But  we  recall  that  in  language  these  permutations 
are  at  the  mercy  of  conflicting  analogical  influences,  and  this 
suffices  to  show  that  such  rules  are  always  precarious  and  fit  per- 
fectly the  definition  of  synchronic  law. 

Sometimes  the  phonetic  cause  of  the  alternation  is  still  evident. 
In  Old  High  German,  for  instance,  the  pairs  cited  on  page  158  had 
the  forms  gehan:  gibit,  feld:  gcfildi,  etc.  During  that  period  the 
radical  itself  appeared  with  i  instead  of  e  wherever  i  followed  but 
with  e  in  every  other  instance.  The  alternation  of  Latin  facio: 
conficio,  amicus:  inimlcus,  facilis:  difficilis,  etc.,  is  likewise  linked 
to  a  phonic  condition  which  speakers  would  have  expressed  in  this 
way :  the  a  of  such  words  as  facto  and  amicus  alternates  with  i  in 
medial  syllables  of  words  in  the  same  family. 

But  the  foregoing  phonic  oppositions  suggest  exactly  the  same 
observations  as  all  grammatical  laws:  they  are  synchronic.  To  for- 


get  this  is  to  risk  making  the  mistake  in  interpretation  pointed  out 
above  (see  pp.  96  ff.).  Faced  with  a  pair  like  f acid:  conficio,  we  must 
indeed  guard  against  confusing  the  relation  between  these  co- 
existing terms  and  the  relation  that  ties  together  the  successive 
terms  of  the  diachronic  fact  (confacio  — >  conficio) .  We  may  be 
tempted  to  confuse  them  since  the  cause  of  phonetic  differentiation 
is  still  apparent  in  the  pair,  but  the  phonetic  fact  belongs  to  the 
past,  and  for  speakers  there  is  only  a  single  synchronic  opposition. 
All  of  this  confirms  what  was  said  about  the  strictly  grammatical 
nature  of  alternation.  The  word  permutation,  which  is  apt  in  some 
ways,  has  been  used  for  alternation  but  should  be  avoided  for  the 
very  reason  that  it  has  often  been  applied  to  phonetic  changes  and 
suggests  a  false  notion  of  movement  where  there  is  only  a  state. 

6.  Alternation  and  Grammatical  Bond 

We  have  seen  how  phonetic  evolution  may  cause  a  break  in  the 
grammatical  bonds  that  unite  words  by  changing  the  form  of  the 
words.  But  this  is  true  only  of  isolated  pairs  like  maison:  menage, 
Teil:  Drittel,  etc.,  not  of  alternation. 

It  is  obvious  from  the  first  that  any  slightly  regular  phonic  oppo- 
sition of  two  elements  tends  to  establish  a  bond  between  them. 
Wetter  is  instinctively  related  to  wittern  because  speakers  are  ac- 
customed to  seeing  e  alternate  with  i.  As  soon  as  speakers  feel  that 
there  is  a  general  law  governing  a  phonic  opposition,  the  usual 
correspondence  has  all  the  more  reason  for  forcing  itself  on  their 
attention  and  helping  to  tighten  rather  than  loosen  the  gram- 
matical bond.  This  is  how  the  German  ablaut  reinforces  recog- 
nitions of  the  radical  unit  across  vocalic  variations  (see  p.  158). 

The  same  is  true  of  nonsignificant  alternations  that  are  linked 
to  a  mere  phonic  condition.  In  French,  the  prefix  re-  (rependre 
'retake,'  regagner  'regain,'  retoucher  'retouch,'  etc.)  is  reduced  to 
r-  before  a  vowel  (rouvrir  'reopen,'  racheter  'buy  back,'  etc.).  Simi- 
larly, under  the  same  conditions  the  prefix  in-,  still  very  much 
alive  although  of  learned  origin,  has  two  distinct  forms:  e-  (in 
inconnu  'unknown,'  indigne  'unworthy,'  invertebre  'invertebrate,' 
etc.)  and  in-  (in  inavouahle  'inadmissible,'  inutile  'useless,'  in- 
esthetique  'unaesthetic,'  etc.).  In  no  way  does  this  difference  break 


unity  of  conception,  for  meaning  and  function  are  apprehended 
as  identical,  and  language  has  determined  where  it  will  use  one 
form  or  the  other. 

Chapter  IV 

1.  Definition  and  Examples 

That  phonetic  evolution  is  a  disturbing  force  is  now  obvious. 
Wherever  it  does  not  create  alternations,  it  helps  to  loosen  the 
grammatical  bonds  between  words;  the  total  number  of  forms  is 
uselessly  increased ;  the  linguistic  mechanism  is  obscured  and  com- 
plicated to  the  extent  that  the  irregularities  born  of  phonetic 
changes  win  out  over  the  forms  grouped  under  general  patterns;  in 
other  words,  to  the  extent  that  absolute  arbitrariness  wins  out 
over  relative  arbitrariness  (see  p.  1.33). 

Fortunately,  analogy  counterbalances  the  effect  of  phonetic 
transformations.  To  analogy  are  due  all  normal,  nonphonetic 
modifications  of  the  external  side  of  words. 

Analogy  supposes  a  model  and  its  regular  imitation.  An  ana- 
logical form  is  a  form  made  on  the  model  of  one  or  more  other  forms 
in  accordance  with  a  definite  rule. 

The  nominative  form  of  Latin  honor,  for  instance,  is  analogical. 
Speakers  first  said  honos:  honosem,  then  through  rhotacization  of 
the  s,  honos:  honorem.  After  that,  the  radical  had  a  double  form. 
This  duality  was  eliminated  by  the  new  form  honor,  created  on  the 
pattern  of  orator:  ordtorem,  etc.,  through  a  process  which  sub- 
sequently will  be  set  up  as  a  proportion : 

ordtorem:  ordtor  =  honorem:  x 
X  =  honor 

Thus  analogy,  to  offset  the  diversifying  action  of  a  phonetic 
change  {honos:  honorem),  again  unified  the  forms  and  restored 
regularity  (honor:  honorem). 

For  a  long  time  French  speakers  said  il  preuve,  nous  prouvons,  its 


preuvent.  Today  they  say  il  prouve  'he  proves/  Us  prouvent  'they 
prove/  using  forms  that  have  no  phonetic  explanation.  II  aime  'he 
loves'  is  derived  from  Latin  amat  while  nous  aimons  'we  love'  is  the 
analogical  form  for  amons;  speakers  should  also  say  amahle  instead 
of  aimable  'amiable.'  In  Greek,  intervocalic  s  disappeared:  -oso- 
became  -eo-  (cf .  geneos  for  *genesos) .  Still,  intervocalic  s  is  found  in 
the  future  and  aorist  tenses  with  s.  In  German,  Gast:  Gdste,  Balg: 
Bdlge,  etc.  are  phonetic,  but  Kranz:  Kranze  (previously  kranz: 
kranza),  Hals:  Hdlse  (previously  halsa),  etc.  are  due  to  imitation. 

Analogy  favors  regularity  and  tends  to  unify  structural  and  in- 
flectional procedures.  But  it  is  capricious;  beside  Kranz:  Kranze, 
etc.,  stand  Tag:  Tage,  Salz:  Salze,  etc.,  which  for  one  reason  or 
another  have  resisted  analogy.  Thus  we  cannot  say  beforehand 
how  far  imitation  of  a  model  will  go  or  which  types  will  bring  it 
about.  The  most  numerous  forms  do  not  necessarily  unleash 
analogy.  The  Greek  perfect  has  the  active  forms  pepheiiga,  pephe- 
ugas,  pepheugamen,  but  all  the  middle  forms  are  inflected  without 
a:  pephugmai,  pephugmetha,  etc.,  and  the  language  of  Homer  shows 
that  the  a  was  formerly  missing  in  the  plural  and  in  the  dual  of  the 
active  (cf.  idmen,  eikion,  etc.).  Analogy  started  solely  from  the 
first  person  singular  of  the  active  and  won  over  almost  the  whole 
paradigm  of  the  perfect  indicative.  This  development  is  also  note- 
worthy because  here  analogy  attached  -a-,  originally  an  inflec- 
tional element,  to  the  radical,  forming  pepheuga-men.  The  reverse — 
attaching  the  radical  element  to  the  suffix — is  much  more  common 
(see  p.  170). 

Two  or  three  words  often  suffice  to  create  a  general  form  such 
as  an  inflectional  ending.  In  Old  High  German,  weak  verbs  like 
haben,  lohon,  etc.  had  an  -m  in  the  first  person  singular  of  the 
present :  hahem,  lohom,  etc.  The  -m  derives  from  a  few  verbs  similar 
to  -^mi  verbs  in  Greek  (bim,  *tdm,  gom,  tuom),  which  by  themselves 
forced  the  ending  on  the  whole  weak  conjugation.  Notice  that  here 
analogy  did  not  eliminate  a  phonetic  difference  but  generalized  a 
formative  method. 

2.  Analogical  Phenomena  Are  Not  Changes 

The  first  linguists  did  not  understand  the  nature  of  the  phe- 
nomenon of  analogy,  which  they  called  "false  analogy."  They 


thought  that  in  inventing  honor,  Latin  "had  made  a  mistake" 
concerning  the  prototype  honos.  For  them,  everything  that  de- 
viated from  the  original  state  was  an  irregularity,  a  distortion  of 
an  ideal  form.  The  fact  is  that,  through  an  illusion  characteristic 
of  their  time,  they  saw  in  the  original  state  of  the  language  some- 
thing superior  and  perfect,  with  the  result  that  they  did  not  even 
ask  themselves  whether  this  state  had  been  preceded  by  another. 
Every  hberty  taken  with  respect  to  this  state  was  then  an  anomaly. 
The  neogrammarian  school  was  the  first  to  assign  analogy  to  its 
proper  place  by  showing  that  it  is,  along  with  phonetic  changes,  the 
prime  force  in  the  evolution  of  languages,  the  procedure  through 
which  languages  pass  from  one  state  of  organization  to  another. 

But  exactly  what  are  analogical  phenomena?  People  generally 
think  of  them  as  changes.  But  are  they? 

Every  analogical  fact  is  a  play  with  a  cast  of  three:  (1)  the 
traditional,  legitimate  heir  (e.g.  honos) ;  (2)  the  rival  (honor) ;  and 
(3)  a  collective  character  made  up  of  the  forms  that  created  the 
rival  (honorem,  orator,  ordtorem,  etc.).  One  might  readily  suppose 
that  honor  is  a  modification,  a  "metaplasm,"  of  honos  and  say  that 
it  drew  most  of  its  substance  from  honos.  But  the  only  form  that 
had  no  part  in  the  production  of  honor  is  this  very  honos! 

The  phenomenon  of  analogy  may  be  pictured  by  the  diagram: 


Honos  honorem, 

(which  plays         orator,  oratorem,  etc.  honor 

no  part)  (productive  group) 

Here  we  obviously  have  a  "paraplasm,"  the  installation  of  a 
rival  beside  a  traditional  form — in  short,  a  creation.  Whereas  pho- 
netic change  introduces  nothing  new  without  annulling  what  has 
preceded  it  (honorem  replaces  honosem),  the  analogical  form  does 
not  necessarily  entail  the  disappearance  of  its  double.  Honor  and 
honos  coexisted  for  a  time  and  were  used  interchangeably.  Still, 
since  language  is  reluctant  to  keep  two  signifiers  for  a  single  idea, 
the  original  form,  which  is  less  regular,  generally  falls  into  disuse 
and  disappears.  The  result  is  what  gives  the  impression  of  a  trans- 
formation. Once  analogy  has  completed  its  work,  the  opposition 


between  the  old  state  {honos:  honorem)  and  the  new  {honor: 
honor  em)  is  apparently  the  same  as  the  opposition  that  results 
from  the  evolution  of  sounds.  At  the  moment  when  honor  was  born, 
however,  nothing  was  changed  since  honor  replaced  nothing;  nor 
is  the  disappearance  of  honos  a  change,  for  this  phenomenon  is 
independent  of  the  first.  Wherever  we  can  follow  the  course  of 
linguistic  events,  we  see  that  analogical  innovation  and  the  elimi- 
nation of  the  older  form  are  two  distinct  things,  and  that  nowhere 
do  we  come  upon  a  transformation. 

So  little  does  analogy  have  the  characteristic  of  replacing  one 
form  by  another  that  it  often  produces  forms  which  replace  nothing 
at  all.  German  can  make  a  diminutive  in  -chen  from  any  sub- 
stantive with  a  concrete  meaning;  if  the  form  Elefantchen  were 
introduced  into  the  language,  it  would  supplant  nothing  that 
already  exists.  Similarly  in  French,  on  the  model  of  pension  pen- 
sion': pensionnaire  'pensionary,'  reaction,  'reaction':  reactionnaire 
'reactionary,'  etc.,  someone  might  create  interventionnaire,  repres- 
sionnaire,  etc.,  meaning  'one  who  favors  intervention,'  'one  who 
favors  repression,'  etc.  The  process  is  evidently  the  same  as  the 
one  that  engendered  honor;  both  recall  the  same  formula : 

reaction:  reactionnaire  =  repression:  x 

X  =  repressionnaire 

In  neither  case  is  there  the  slightest  pretext  for  speaking  of  change ; 
repressionnaire  replaces  nothing.  Another  example:  some  French 
speakers  use  the  analogical  form  finaux  instead  of  finals,  which  is 
more  common;  someone  might  coin  the  &dieciive  firmamental  and 
give  it  the  plural  form  firmamentaux.  Should  we  say  that  there  is 
change  in  finaux  and  creation  in  firmamentaux?  In  both  cases  there 
is  creation.  On  the  pattern  of  mur  'wall' :  enmurer  'wall  in,'  speakers 
formed  tour  'turn':  entourer  'surround,'  and  jour  'light':  ajourer 
'open'  (in  un  travail  ajoure  'work  that  admits  light,  i.e.  lacework,' 
etc.).  These  rather  recent  derivatives  seem  to  be  creations.  But  if 
I  notice  that  entorner  and  ajorner,  built  on  torn  and  jorn,  were  used 
during  an  earlier  period,  must  I  change  my  mind  and  say  that 
entourer  and  ajourer  are  modifications  of  the  older  words?  The 
illusion  of  analogical  change  comes  from  setting  up  a  relation  be- 
tween the  new  form  and  the  one  replaced  by  it.  But  this  is  a  mis- 


take  since  formations  classed  as  changes  (like  honor)  are  basically 
the  same  as  those  I  call  creations  (like  repressionnaire) . 

3.  Analogy  as  a  Creative  Force  in  Language 

When,  after  seeing  what  analogy  is  not,  we  begin  to  study  it  for 
what  it  is,  we  find  that  it  seems  very  simply  to  blend  with  the 
principle  of  linguistic  creativity  in  general.  What  is  that  principle? 

Analogy  is  psychological,  but  this  does  not  suffice  to  separate 
it  from  phonetic  phenomena,  for  they  may  also  be  considered 
psychological  (see  p.  151).  We  must  go  further  and  say  that  anal- 
ogy is  grammatical.  It  supposes  awareness  and  understanding 
of  a  relation  between  forms.  Meaning  plays  no  part  in  phonetic 
changes,  but  it  must  intervene  in  analogy. 

As  far  as  we  can  tell,  neither  comparison  with  other  forms  nor 
meaning  had  anything  to  do  with  the  passing  from  intervocalic  s 
to  r  in  Latin.  The  skeleton  of  the  form  honosem  passed  to  honor  em. 
Other  forms  must  be  introduced  to  account  for  the  appearance  of 
honor  beside  honos.  This  is  shown  by  the  proportion: 

drat  or  em:  orator  =  honor  em:  x 
X  =  honor 

The  new  combination  would  have  no  basis  if  the  mind  did  not 
associate  its  forms  through  their  meanings. 

Analogy  is  grammatical  throughout,  but  let  us  hasten  to  add 
that  its  end  result — creation — belongs  at  first  only  to  speaking.  It 
is  the  chance  product  of  an  isolated  speaker.  Here,  at  the  very 
fringe  of  language,  is  where  the  phenomenon  must  first  be  sought. 
Still,  two  things  must  be  kept  apart:  (1)  awareness  of  the  relation 
that  ties  together  the  productive  forms;  and  (2)  the  result  sug- 
gested by  the  comparison,  the  form  improvised  by  the  speaker  to 
express  his  thought.  Only  the  result  belongs  to  speaking. 

Analogy,  then,  is  one  more  lesson  in  separating  language  from 
speaking  (see  pp.  17  ff.).  It  shows  us  that  the  second  depends  on 
the  first,  and  it  points  to  the  essence  of  the  linguistic  mechanism  as 
described  on  page  130.  Any  creation  must  be  preceded  by  an  un- 
conscious comparison  of  the  materials  deposited  in  the  storehouse 
of  language,  where  productive  forms  are  arranged  according  to 
their  syntagmatic  and  associative  relations. 


A  major  part  of  the  analogical  phenomenon  is  therefore  com- 
pleted before  the  new  form  appears.  Speech  is  continuously  en- 
gaged in  decomposing  its  units,  and  this  activity  contains  not  only 
every  possibility  of  effective  talk,  but  every  possibility  of  ana- 
logical formation.  It  is  wrong  to  suppose  that  the  productive  proc- 
ess is  at  work  only  when  the  new  formation  actually  occurs.  The 
elements  were  already  there.  A  newly  formed  word  like  in-decor- 
ahle  already  has  a  potential  existence  in  language;  all  its  elements 
are  found  in  syntagms  like  decor-er  'decorate,'  decor-ation  'decor- 
ation,' pardonn-able  'pardonable,'  mani-able  'manageable':  in- 
connu  'unknown,'  in-sense  'insane,'  etc.,  and  the  final  step  of 
realizing  it  in  speaking  is  a  small  matter  in  comparison  with  the 
build-up  of  forces  that  makes  it  possible. 

In  short  analogy,  considered  by  itself,  is  only  one  side  of  the 
phenomenon  of  interpretation,  one  manifestation  of  the  general 
activity  that  singles  out  units  for  subsequent  use.  That  is  why  I 
say  that  analogy  is  entirely  grammatical  and  sjmchronic. 

The  grammatical  and  synchronic  character  of  analogy  suggests 
two  observations  that  confirm  my  views  on  absolute  and  relative 
arbitrariness  (see  pp.  131  ff.). 

1)  Words  can  be  rated  for  capacity  to  engender  other  words  to 
the  extent  to  which  they  themselves  are  decomposable.  Simple 
words  are  by  definition  unproductive  (cf.  French  magasin  'ware- 
house,' arbre  'tree,'  racine  'root,'  etc.).  Magasinier  'warehouse- 
keeper'  was  not  engendered  by  magasin.  It  was  formed  on  the  pat- 
tern of  prisonier  'prisoner':  prison  'prison,'  etc.  In  the  same  way 
emmagisiner  'to  warehouse'  owes  its  existence  to  the  analogy  of 
enmailloter  'swathe,'  encadrer  'frame/  encapuchonner  'put  on  a 
cowl,'  etc.,  which  contain  maillot  'swaddling-clothes,'  cadre  'frame,' 
capuchon  'cowl,'  etc. 

Each  language  then  has  both  productive  and  sterile  words,  in 
varying  proportions.  This  takes  us  back  to  the  distinction  between 
"lexicological"  and  "grammatical"  languages  (see  p.  133).  In 
Chinese,  most  words  are  not  decomposable;  in  an  artificial  lan- 
guage, however,  almost  all  words  are.  An  Esperantist  has  un- 
limited freedom  to  build  new  words  on  a  given  root. 

2)  We  have  seen  (p.  161)  that  any  analogical  creation  may  be 
pictured  as  similar  to  a  proportion.  This  formula  is  frequently  used 


to  explain  the  phenomenon  of  analogical  creation  itself,  but  we 
have  sought  its  explanation  in  the  analysis  and  reconstruction  of 
elements  furnished  by  language. 

There  is  a  conflict  between  the  two  notions.  If  proportion  is  a 
satisfactory  explanation,  why  posit  an  analysis  of  elements?  To 
form  indecorable,  there  is  no  point  in  extracting  its  elements  (m- 
decor-able).  All  we  need  do  is  to  take  the  whole  and  put  it  in  the 
equation : 

pardonner:  impardonnable,  etc.  =  decorer:  x 

X  =  indecorable 

Here,  no  compUcated  operation  such  as  the  grammarian's  con- 
scious analysis  is  presumed  on  the  part  of  the  speaker.  In  Krantz: 
Krdnze,  modeled  on  Gast:  Gdste  and  the  like,  decomposition  seems 
less  probable  than  proportion  since  the  radical  of  the  model  may 
be  either  Gast-  or  Gdst-.  A  phonic  characteristic  of  Gdste  might 
simply  have  been  carried  over  to  Kranze. 

Of  the  two  theories,  which  fits  the  facts?  (Bear  in  mind  that 
Kranz  does  not  necessarily  exclude  analysis.  We  have  observed 
alternations  in  roots  and  prefixes,  and  the  feeling  for  alternation 
may  well  exist  alongside  positive  analysis;  see  p.  158.) 

The  two  contrasting  notions  are  reflected  in  two  different  gram- 
matical doctrines.  European  grammars  work  with  proportion ;  they 
explain  the  formation  of  the  German  preterite,  for  example,  by 
starting  from  whole  words.  On  the  model  of  setzen:  setzte  the  pupil 
is  told  to  form  the  preterite  of  lachen,  etc.  Against  this,  Hindu 
grammar  would  study  roots  (setz-,  lack-,  etc.)  in  one  chapter  and 
preterite  endings  {-te,  etc.)  in  another.  The  elements  that  result 
from  analysis  would  be  given,  and  from  these  elements  whole  words 
would  have  to  be  reconstructed.  In  every  Sanskrit  dictionary, 
verbs  are  arranged  in  the  order  assigned  to  them  by  their  roots. 

Theoreticians  of  grammar  will  incline  toward  whichever  method 
is  predominant  in  their  linguistic  group. 

Old  Latin  apparently  favors  the  analytical  procedure.  Here  is 
obvious  proof :  quantity  is  not  the  same  in  f actus  and  actus  despite 
fdcio  and  ago;  we  must  assume  that  actus  goes  back  to  *dgtos  and 
attribute  lengthening  of  the  vowel  to  the  voiced  consonant  that 
followed;  this  hypothesis  is  fully  confirmed  by  the  Romance  Ian- 


guages.  The  opposition  specio:  spectus  against  tego:  tectus  is  reflected 
in  French  depit  'despite'  (=  despedus)  and  toil  'roof  (=  tectum); 
cf.  conficio:  confectus  (French  confit  'candied')  against  rego:  rectus 
{dlrectus  —^  French  droit  'straight').  But  *agtos,  *tegtos,  *regtos 
were  not  inherited  from  Proto-Indo-European,  which  certainly  had 
*aktos,  *tektos,  etc.;  prehistoric  Latin  introduced  them,  and  this 
despite  the  diflSculty  of  pronouncing  a  voiced  consonant  before  a 
voiceless  one.  This  was  made  possible  only  by  acute  awareness  of 
the  radical  units  ag-,  teg-,  reg-.  The  f eehng  for  word-parts  (radicals, 
suffixes,  etc.)  and  their  arrangement  was  therefore  strong  in  Old 
Latin.  In  all  probability  the  feeling  is  not  so  acute  in  modern  lan- 
guages but  is  stronger  in  German  than  in  French  (see  p.  186  f.). 

Chapter  V 


L  How  an  Analogical  Innovation  Enters  Language 

Nothing  enters  language  without  having  been  tested  in  speaking, 
and  every  evolutionary  phenomenon  has  its  roots  in  the  individual. 
This  principle,  which  was  stated  previously  (see  p.  98),  applies 
particularly  to  analogical  innovations.  Before  honor  could  become 
a  rival  strong  enough  to  replace  honos,  one  speaker  had  to  coin  the 
new  word,  then  others  had  to  imitate  and  repeat  it  until  it  forced 
itself  into  standard  usage. 

But  not  every  analogical  innovation  is  so  fortunate.  Abortive 
combinations  that  language  will  probably  never  adopt  are  always 
at  hand.  Children,  because  they  are  not  well  acquainted  with 
standard  usage  and  are  not  yet  bound  by  it,  clutter  their  speech 
with  them:  in  French  they  say  viendre  for  venir  'come,'  mouru  for 
mort  'dead,'  etc.  But  adults  use  them  too.  For  instance,  many  peo- 
ple say  traisait  (which,  incidentally,  is  found  in  the  writings  of 
Rousseau)  instead  of  tray  ait  '(he)  milked.'  All  such  innovations 
are  perfectly  regular;  they  are  explained  in  the  same  way  as  those 


that  language  has  accepted;  viendre,  for  example,  stems  from  the 
proportion : 

eteindrai:  iteindre  =  viendrai:  x 
X  =  viendre 

and  traisait  was  formed  on  the  model  of  plaire  'please':  plaisait 
*(he)  pleased,'  etc. 

Language  retains  only  a  minimal  part  of  the  creations  of  speak- 
ing, but  those  that  endure  are  numerous  enough  to  change  com- 
pletely the  appearance  of  its  vocabulary  and  grammar  from  one 
period  to  the  next. 

From  what  was  said  in  the  preceding  chapter,  it  is  evident  that 
analogy  by  itself  could  not  be  a  force  in  evolution,  and  that  the 
constant  substitution  of  new  forms  for  old  ones  is  one  of  the  most 
striking  features  in  the  transformation  of  languages.  Each  time  a 
new  formation  becomes  definitely  installed  and  eliminates  its  rival, 
something  is  actually  created  and  something  else  abandoned,  with 
the  result  that  analogy  occupies  a  preponderant  place  in  the  theory 
of  evolution. 

This  is  the  point  that  I  should  like  to  emphasize. 

2.  Analogical  Innovations  as  Symptoms  of  Changes  in  Interpretation 
Language  never  stops  interpreting  and  decomposing  its  units. 
But  why  does  interpretation  vary  constantly  from  one  generation 
to  the  next?  The  cause  of  change  must  be  sought  in  the  great  mass 
of  forces  that  constantly  threaten  the  analysis  adopted  in  a 
particular  language-state.  I  shall  recall  a  few  of  them. 

The  first  and  most  important  force  is  phonetic  evolution  (see 
Chapter  II) .  By  making  some  analyses  ambiguous  and  others  im- 
possible, phonetic  changes  affect  both  the  conditions  and  the 
results  of  decomposition,  thereby  shifting  the  boundaries  and 
changing  the  nature  of  units  (see  p.  141  concerning  compounds 
like  heta-hUs  and  redo-lich,  and  p.  155  concerning  noun  inflection 
in  Proto-Indo-European). 

In  addition  to  the  phonetic  fact  there  is  agglutination  (to  be 
discussed  later),  which  welds  a  combination  of  elements  into  one 
unit,  and  every  imaginable  circumstance  which,  though  external, 


may  modify  the  analysis  of  words.  For  it  is  obvious  that  analysis, 
because  it  results  from  a  set  of  comparisons,  depends  constantly  on 
the  associative  environment  of  the  term.  The  Proto-Indo-European 
superlative  *swdd-is-to-s  contained  two  independent  suffixes,  -is-, 
which  carried  the  idea  of  comparative  degree  (cf .  Latin  mag-is)  and 
-to~,  which  designated  the  definite  place  of  an  object  in  a  series  (cf, 
Greek  trl-to-s  'third').  The  two  prefixes  were  agglutinated  (cf. 
Greek  hed-isto-s,  or  rather  hed-ist-os).  But  agglutination  was  in 
turn  greatly  aided  by  a  fact  unrelated  to  the  concept  of  the  su- 
perlative degree:  comparatives  in  is-  had  dropped  out  of  usage, 
having  been  supplanted  by  formations  in  -jos;  since  -is-  was  no 
longer  recognized  as  an  independent  element,  it  was  no  longer 
singled  out  in  -isto-. 

We  note  in  passing  the  general  tendency  to  shorten  the  radical 
in  favor  of  the  formative  element,  especially  when  the  former  ends 
in  a  vowel.  Thus  the  Latin  suffix  -tat-  (veri-tdt-em  for  vero-tdt-em, 
cf.  Greek  deino-tet-a)  took  over  the  i  of  the  theme,  giving  the 
analysis  ver-itdt-em;  in  the  same  way  Romd-nus,  Albd-nus  (cf .  aenus 
for  *aesno-s)  became  Rom-dnus,  etc. 

Changes  in  interpretation,  no  matter  how  they  start,  always 
become  apparent  through  the  existence  of  analogical  forms.  Indeed, 
if  living  units  perceived  by  speakers  at  a  particular  moment  can  by 
themselves  give  birth  to  analogical  formations,  every  definite  re- 
distribution of  units  also  implies  a  possible  expansion  of  their  use. 
Analogy  is  therefore  proof  positive  that  a  formative  element  exists 
at  a  given  moment  as  a  significant  unit.  Meridiondlis  (Lactantius) 
for  merldidlis  shows  that  the  division  was  septentri-ondlis,  regi- 
ondlis,  and  to  prove  that  the  suffix  -tat  had  been  enlarged  by  an  i 
element  borrowed  from  the  radical,  we  need  only  cite  celer-itdtem; 
pdg-dnus,  built  on  pdg-us,  suffices  to  show  how  Latin  speakers 
analyzed  Rom-dnus;  and  the  analysis  of  redlich  (see  p.  141)  is  con- 
firmed by  the  existence  of  sterhlich,  formed  with  a  verbal  root. 

A  particularly  unusual  example  will  show  how  analogy  works  out 
new  units  from  period  to  period.  In  Modem  French,  somnolent 
'sleepy'  is  analyzed  somnol-ent,  as  if  it  were  a  present  participle. 
Proof  of  this  is  the  existence  of  the  verb  somnoler  'be  sleepy.'  But 
in  Latin  the  division  was  somno-lentus,  like  succu-lentus,  etc.,  and 


before  that  it  was  somn-olentus  'smelling  of  sleep,'  from  olere,  as  in 
vln-olentus  'smelling  of  wine.' 

The  most  obvious  and  important  effect  of  analogy  is  thus  the 
substituting  of  more  regular  forms  composed  of  living  elements  for 
older  irregular  and  obsolescent  forms. 

Doubtless  things  do  not  always  run  so  smoothly.  The  functioning 
of  language  is  disturbed  by  many  hesitations,  approximations,  and 
semianalyses.  At  no  time  does  an  idiom  have  a  perfectly  stable 
system  of  units.  From  what  was  said  about  the  inflection  of  *ekwos 
against  *pods,  it  is  obvious  that  imperfect  analyses  sometimes  lead 
to  muddled  analogical  creations.  The  Proto-Indo-European  forms 
*geus-etai,  *gus-tos,  *gus-tis  allow  us  to  single  out  the  root  *geus-, 
gus-.  But  intervocalic  s  fell  in  Greek,  and  the  analysis  of  geuomai, 
geustos  was  accordingly  beclouded.  Fluctuation  resulted,  and  the 
root  singled  out  was  sometimes  geus-,  sometimes  geu-.  Analogy  in 
turn  bears  witness  to  this  fluctuation,  for  even  roots  in  eu-  take 
final  -s  (e.g.  pneu-,  pneuma,  and  the  verbal  adjective  pneus-tos). 

But  analogy  influences  language  even  when  there  is  groping  and 
hesitation.  For  analog}'-,  though  not  an  evolutionary  fact  in  itself, 
usually  reflects  the  changes  that  have  affected  the  functioning  of 
language  and  sanctions  them  through  new  combinations.  It  col- 
laborates efficiently  with  all  the  forces  that  constantly  modify  the 
architecture  of  an  idiom  and  is  in  this  way  a  powerful  force  in 

3.  Analogy  as  a  Renovating  and  Conservative  Force 

One  is  sometimes  tempted  to  ask  whether  analogy  actually  has 
the  importance  attributed  to  it  here  and  whether  its  action  is  as 
far-reaching  as  that  of  phonetic  changes.  As  a  matter  of  fact,  the 
history  of  each  language  discloses  a  motley  accumulation  of  ana- 
logical facts.  Collectively,  these  continuous  reshufflings  play  an 
even  more  important  part  in  the  evolution  of  language  than  do 
sound  changes. 

But  one  thing  in  particular  interests  the  linguist.  In  the 
enormous  mass  of  analogical  phenomena  built  up  through  cen- 
turies of  evolution,  almost  all  elements  are  preserved;  they  are  only 
distributed  differently.  Analogical  innovations  are  more  apparent 


than  real.  Language  is  a  garment  covered  with  patches  cut  from 
its  own  cloth.  Four-fifths  of  French  is  Proto-Indo-European  if  we 
think  of  the  substance  that  constitutes  sentences,  but  the  words 
that  have  been  transmitted  in  their  totahty  without  analogical 
change  from  the  mother  language  to  Modern  French  would  occupy 
less  than  the  space  of  one  page  (e.g.  est  'is'  =  *esti,  numbers,  words 
like  ours  'bear,'  nez  'nose,'  pere  'father,'  chien  'dog,'  etc.).  The  vast 
majority  of  words  are,  in  one  way  or  another,  new  combinations  of 
phonic  elements  torn  from  older  forms.  In  this  sense  analogy,  for 
the  very  reason  that  it  always  uses  old  material  for  its  innovations, 
is  remarkably  conservative. 

But  analogy  has  an  equally  important  role  as  a  conservative 
force  pure  and  simple.  It  intervenes  not  only  when  old  materials 
are  redistributed  in  new  units  but  also  when  forms  remain  un- 
changed. To  realize  this,  we  need  only  recall  that  analogical  cre- 
ation and  the  mechanism  of  speech  have  a  common  basis  (see 
p.  165). 

Latin  agunt  was  transmitted  almost  intact  from  the  prehistoric 
period  (when  people  said  *agonti)  until  the  beginning  of  the  Ro- 
mance period.  During  that  span  of  time  successive  generations 
used  the  form  over  and  over  without  there  being  a  rival  form  to 
replace  it.  Here  analogy  played  a  part  in  the  retention  of  the  form. 
The  stability  of  agunt  is  just  as  much  the  work  of  analogy  as  is  any 
innovation.  Agunt  is  integrated  in  a  system;  it  is  supported  by 
forms  like  dicunt  and  legunt  as  well  as  by  agimus,  agitis,  and  the 
like.  Outside  this  frame,  agunt  might  easily  have  been  replaced  by 
a  form  made  up  of  new  elements.  What  was  transmitted  was  not 
agunt  but  ag-unt.  The  form  did  not  change  because  ag-  and  -wn< 
regularly  appeared  in  other  series,  and  the  support  of  these  forms 
preserved  agunt  from  start  to  finish.  Compare  also  sex-tus,  which 
is  supported  by  two  compact  series:  sex,  sex-aginta,  etc.  on  the  one 
hand  and  quar-tus,  quin-tus,  etc.  on  the  other. 

Forms  are  then  preserved  because  they  are  constantly  renewed 
by  analogy.  A  word  is  apprehended  simultaneously  as  a  unit  and 
as  a  syntagm,  and  is  preserved  to  the  extent  that  its  elements  do 
not  change.  Conversely,  the  existence  of  the  form  is  threatened 
only  to  the  extent  that  its  elements  disappear  from  usage.  Con- 
sider what  is  happening  to  French  dites  '(you)  say'  and  faites  '(you) 


do,'  which  are  direct  descendants  of  Latin  dic-itis  and  fac-itis.  Be- 
cause they  have  no  support  from  present-day  verbal  inflection, 
language  is  trying  to  replace  them.  Disez,  faisez  (on  the  pattern  of 
plaisez  'please,'  lisez  'read,'  etc.)  are  heard  today,  and  the  new  end- 
ings are  already  common  in  most  compounds  {contredisez  'contra- 
dict,' etc.). 

The  only  forms  left  untouched  by  analogy  are  of  course  isolated 
words  like  proper  nouns,  especially  place  names  (cf .  Paris,  Geneva, 
Agen,  etc.),  which  allow  no  analysis  and  consequently  no  interpre- 
tation of  their  elements.  No  rival  creation  springs  up  beside  them. 

It  follows  that  a  form  may  be  preserved  for  either  of  two  dia- 
metrically opposed  reasons:  complete  isolation  or  complete  in- 
tegration in  a  system  that  has  kept  the  basic  parts  of  the  word 
intact  and  that  always  comes  to  its  rescue.  It  is  within  the  inter- 
mediate group  of  forms  not  supported  firmly  enough  by  their 
environment  that  innovating  analogy  may  unfold  its  effects. 

But  whether  we  deal  with  the  preservation  of  a  form  composed 
of  several  elements  or  a  redistribution  of  linguistic  material  in  new 
constructions,  analogy  is  there.  It  always  plays  an  important  role. 

Chapter  VI 

We  sometimes  mangle  words  that  have  unfamiliar  forms  and  mean- 
ings, and  usage  sometimes  sanctions  these  deformations.  In  this 
way  Old  French  coute-pointe  (from  coute,  variant  of  couette  'cover' 
and  pointe,  past  participle  of  poindre  'quilt')  was  changed  to  coute- 
pointe  'counterpane,'  as  if  formed  from  the  adjective  court  'short' 
and  the  noun  pointe  'point.'  *  Such  innovations,  no  matter  how  odd 
they  may  seem,  are  not  due  entirely  to  chance;  they  are  crude  at- 
tempts to  explain  refractory  words  by  relating  them  to  something 

At  first  blush  this  phenomenon,  called  folk  etymology,  can 
*  Cf.  Old  English  scam-faest  'confirmed  in  shame.'  In  early  Modern  English 
this  became  shame-fast,  then  shame-faced.  [Tr.] 


hardly  be  distinguished  from  analogy.  When  a  speaker  forgets  that 
French  surdite  'deafness'  exists  and  coins  analogical  sourdite,^  the 
result  is  the  same  as  if  he  had  misunderstood  surdiie  and  deformed 
it  through  remembrance  of  the  adjective  sourd  'deaf;  the  only 
apparent  difference  is  that  analogical  constructions  are  rational 
while  folk  etymology  works  somewhat  haphazardly  and  results 
only  in  absurdities. 

But  this  difference,  which  concerns  only  the  results,  is  not  basic. 
Their  basic  dissimilarity  goes  much  deeper.  In  order  to  see  what  it 
is,  let  us  begin  by  citing  a  few  examples  of  the  main  types  of  folk 

First  come  words  that  receive  new  interpretations  with  no  cor- 
responding change  of  form.  In  German,  durchblduen  'thrash 
soundly'  goes  back  etymologically  to  hliuwan  'flog'  but  is  associated 
with  hlau  'blue'  because  of  the  "blues"  produced  by  flogging.  In 
the  Middle  Ages  German  borrowed  adventure  'adventure'  from 
French  and  formed  regularly  dhentiire,  Ahenteuer;  without  defor- 
mation the  word  was  associated  with  Abend  ("a  story  related  in 
the  evening") ;  the  result  was  that  during  the  eighteenth  century 
the  word  was  written  Abendteuer.  Old  French  soufraite  'privation' 
(=  suffrada  from  subfrangere)  produced  the  adjective  souffreteux 
'sickly,'  now  associated  with  souffrir  'suffer,'  with  which  it  has 
nothing  in  common.^  French  lais  is  the  noun  form  of  laisser 
'leave'  but  is  associated  nowadays  with  leguer  'bequeath'  and 
written  legs;  some  people  even  pronounce  it  le-g-s?  This  might 
suggest  that  a  change  of  form  resulted  from  the  new  interpretation, 
but  the  change  actually  relates  to  the  influence  of  the  written  form 
through  which  people  tried  to  show  their  idea  of  the  origin  of  the 
word  without  changing  its  pronunciation.  Similarly,  French  ho- 
mard  'lobster,'  borrowed  from  Old  Norse  hummor  (cf.  Danish 
hummer),  added  a  final  d  through  analogy  with  French  words  in 
-ard;  only  here  the  mistake  in  interpretation  that  is  marked  by 
orthography  affects  the  ending,  which  was  confused  with  a  common 
sufl&x  (cf.  bavard  'chatterbox,'  etc.). 

But  people  more  often  deform  words  in  order  to  adapt  them  to 

^  Cf.  English  pronounciation  against  pronunciation.  [Tr.] 
*Cf.  English  liquorice  (from  Latin  liquiritia),  which  has  only  a  graphic 
relation  to  liquor.  [Tr.] 

^  Cf.  English  gooseberry  (from  French  groseiUe).  [Tr.] 


the  elements  which  they  think  they  recognize  in  them.  German 
Sauerkraut  became  choucroute  (chou  'cabbage'  and  croute  'crust') 
in  French.  In  German,  dromeddrius  became  trampeltier  'animal 
that  paws'  in  a  new  compound  which  includes  existing  words, 
trampeln  and  Tier.  Old  High  German  changed  Latin  margarita  to 
mari-greos  'sea-pebble'  by  combining  two  known  words. 

A  last  example,  especially  instructive:  Latin  carbunculus  'small 
piece  of  coal'  became  Karfunkel  (through  association  with  funkeln 
'glow')  in  German  and  escarhoucle  'carbuncle'  (associated  with 
boucle  'buckle,  ring')  in  French.  Calfeter,  calfetrer  became  calfeutrer 
'chink'  in  French  under  the  influence  oi  feutre  'felt.'^  What  strikes 
one  at  the  outset  is  that  each  of  the  examples  contains,  beside  an 
intelligible  element  that  occurs  in  other  contexts,  one  part  that 
stands  for  nothing  that  has  previously  existed  (Kar-,  escar-,  col-). 
But  it  would  be  a  mistake  to  think  that  the  elements  are  partly 
creations,  that  something  new  appeared  as  a  result  of  the  phe- 
nomenon. The  reverse  is  true:  interpretation  could  not  touch  the 
parts  (Kar-,  escar-,  cat-).  We  might  say  that  they  are  parts  of  folk 
etymologies  that  stopped  at  the  half-way  point.  Karfunkel  is  in  the 
same  class  as  Ahenteuer  (if  -teuer  is  considered  an  unexplained  resi- 
due) ;  it  is  also  comparable  to  homard,  where  horn-  makes  no  sense 
by  itself. 

Thus  the  degree  of  deformation  does  not  create  radical  differ- 
ences between  words  corrupted  by  folk  etymology ;  all  these  words 
are  pure  and  simple  interpretations  of  misunderstood  forms  in 
terms  of  known  forms. 

Now  we  see  how  etymology  resembles  analogy,  yet  differs  from 

The  two  phenomena  have  only  one  common  characteristic :  peo- 
ple use  significant  elements  provided  by  language  in  both,  but  the 
two  are  diametrically  opposed  in  everything  else.  Analogy  always 
implies  the  forgetting  of  the  older  forms ;  no  analysis  of  the  older 
form  il  trayait  is  at  the  base  of  the  analogical  form  il  traisait  (see 
p.  168).  The  older  form  must  even  be  forgotten  before  the  rival  can 
appear.  Analogy  takes  nothing  from  the  substance  of  the  signs  that 
it  replaces.  Against  this,  folk  etymology  is  simply  an  interpretation 
of  the  older  form ;  remembrance  of  the  older  form,  though  muddled, 

*  Cf.  English  crayfish,  derived  from  Old  French  crevice,  f Tr.] 


is  the  starting  point  of  the  deformation  that  it  underwent.  The 
basis  for  analysis  is  remembrance  in  one  instance  and  forgetfulness 
in  the  other,  and  this  difference  is  of  prime  importance. 

Folk  etymology  works  only  under  particular  conditions,  then, 
and  affects  only  rare,  technical,  or  foreign  words  that  speakers 
assimilate  imperfectly.  But  analogy,  a  universal  fact,  belongs  to 
the  normal  functioning  of  language.  These  two  phenomena,  so 
similar  in  some  ways,  are  basically  different.  They  must  be  care- 
fully separated. 

Chapter  VII 

1,  Definition 

The  importance  of  analogy  was  indicated  in  the  last  two  chap- 
ters. Along  with  analogy  there  is  another  force  at  work  in  the  pro- 
duction of  new  units:  agglutination. 

Aside  from  these  two,  no  other  formative  device  amounts  to 
much.  Onomatopoeia  (see  p.  69),  words  formed  consciously  and 
without  recourse  to  analogy  by  an  individual  (e.g.  gas),  and  even 
folk  etymology  are  of  little  or  no  importance. 

Agglutination  is  the  welding  together  of  two  or  more  originally 
distinct  terms  that  frequently  occur  as  a  syntagm  within  the  sen- 
tence into  one  unit  which  is  absolute  or  hard  to  analyze.  Such  is  the 
agglutinative  process.  It  is  a  process,  not  a  procedure,  for  the  latter 
word  implies  wall  or  intention,  and  the  absence  of  will  is  what 
characterizes  agglutination. 

Here  are  some  examples.  French  speakers  first  said  ce  ci,  using 
two  words,  then  ceci  'this' :  a  new  word  was  the  result  even  though 
its  substance  and  constituents  did  not  change.  Compare  also: 
French  tous  jours  'every  day,'  toujours  'always,'  au  jour  d'hui  'on 
today's  day,'  aujoiird'hui  'today,'  desjd  'since  now,'  dejd  'already,' 
vert  jus  'green  juice,'  verjus  'verjuice,  sour  grapes.'  Agglutination 
may  also  weld  together  the  subunits  of  a  word,  as  we  saw  (p.  170) 


in  the  case  of  the  Proto-Indo-European  superlative  *swdd-is-to-s 
and  the  Greek  superlative  hed-isto-s. 

On  closer  examination  we  discern  three  phases  in  the  phe- 
nomenon of  agglutination : 

1)  The  combining  of  several  terms  in  a  syntagm.  The  new 
syntagm  is  like  all  other  syntagms. 

2)  Agglutination  proper,  or  the  synthesizing  of  the  elements  of 
the  syntagm  into  a  new  unit.  Synthesis  takes  place  independently 
through  a  mechanical  tendency;  when  a  compound  concept  is 
expressed  by  a  succession  of  very  common  significant  units,  the 
mind  gives  up  analysis — it  takes  a  short-cut — and  apphes  the  con- 
cept to  the  whole  cluster  of  signs,  which  then  become  a  simple  unit. 

3)  Every  other  change  necessary  to  make  the  old  cluster  of  signs 
more  like  a  simple  word:  unification  of  accent  (vert-jus  — >  verjus), 
special  phonetic  changes,  etc. 

It  is  often  claimed  that  phonetic  and  accentual  changes  (3)  pre- 
cede conceptual  changes  (2),  and  that  semantic  synthesis  is  ex- 
plained through  agglutination  and  material  synthesis.  But  this 
probably  puts  the  cart  before  the  horse.  It  is  quite  likely  that  vert 
jus,  tous  jours,  etc.  became  simple  words  because  they  were  grasped 
as  a  single  idea. 

2.  Agglutination  and  Analogy 

The  contrast  between  analogy  and  agglutination  is  striking: 

1)  In  agglutination  two  or  more  units  are  blended  into  one 
through  synthesis  (e.g.  French  encore  'still'  from  hanc  horam),  or 
two  subunits  become  one  (cf.  hed-isto-s  from  *swad-is-to-s) .  Against 
this,  analogy  starts  from  lesser  units  and  builds  them  into  greater 
units.  To  create  pdg-dnus,  analogy  united  the  radical  pdg-  and  the 
suffix  -anus. 

2)  Agglutination  works  only  in  the  zone  of  syntagms.  It  affects 
only  a  particular  cluster.  It  embraces  nothing  else.  In  contrast, 
analogy  calls  forth  associative  series  as  well  as  syntagms. 

3)  Above  all,  agglutination  is  neither  wilful  nor  active.  I  have 
already  said  that  it  is  a  simple  mechanical  process  in  which  merger 
takes  place  spontaneously.  Analogy,  on  the  contrary,  is  a  pro- 
cedure that  requires  analyses  and  combinations,  intelligent  action, 
and  intention. 


Construction  and  structure  are  often  used  in  discussing  word  for- 
mation, but  their  meaning  differs,  depending  on  whether  they  are 
apphed  to  agglutination  or  to  analogy.  When  applied  to  aggluti- 
nation, they  suggest  that  the  elements  in  contact  in  a  syntagm 
slowly  set,  i.e.  are  sjmthesized  to  such  an  extent  that  their  original 
components  are  wiped  out  completely.  But  when  applied  to 
analogy,  construction  means  the  arrangement  obtained  in  one 
swoop,  in  an  act  of  speaking,  by  the  reuniting  of  a  certain  number 
of  elements  borrowed  from  different  associative  series. 

The  importance  of  separating  the  two  formative  methods  is 
obvious.  In  Latin,  for  instance,  possum  is  only  the  welding  to- 
gether of  two  words,  potis  and  sum  'I  am  the  master' :  it  is  an  ag- 
glutinate word.  In  contrast,  signifer,  agricola,  etc.,  are  products 
of  analogy,  constructions  based  on  models  furnished  by  the  lan- 
guage. Only  analogical  creations  may  be  named  compounds  or 
derivatives. '^ 

Often  it  is  difficult  to  say  whether  an  analyzable  form  arose 
through  agglutination  or  as  an  analogical  construction.  Linguists 
have  discussed  endlessly  the  question  of  the  Proto-Indo-European 
forms  *es-mi,  *es-ti,  *ed-mi,  etc.  Were  the  elements  es-,  ed-,  etc.  real 
words  during  a  very  old  period,  and  were  they  later  agglutinated 
with  other  words  (mi,  ti,  etc.)?  Or  are  *es-mi,  *es-ti,  etc.  the  result 
of  combinations  of  elements  drawn  from  other  similar  complex 
units?  In  the  latter  case,  agglutination  would  antedate  the  for- 
mation of  inflectional  endings  in  Proto-Indo-European.  In  the 

'  This  amounts  to  saying  that  the  two  phenomena  act  jointly  in  the  history 
of  language.  But  agglutination  always  occurs  first  and  is  what  furnishes 
models  for  analogy.  For  instance,  the  type  of  compound  that  gave  hippo- 
dromo-s,  etc.  in  Greek  started  through  partial  agglutination  at  a  period  when 
inflectional  endings  were  unknown  in  Proto-Indo-European  {ekwo  dromo  was 
then  equivalent  to  a  compound  like  country  house)  but  through  analogy  be- 
came a  productive  means  of  forming  new  compounds  before  complete  welding 
of  its  elements  occurred.  The  same  is  true  of  the  future  tense  in  French  (Je 
ferai  'I  shall  do,'  etc.),  which  arose  in  Vulgar  Latin  through  agglutination 
of  the  infinitive  with  the  present  tense  of  the  verb  habere  (facere  habed  'I  have 
to  do').  Through  the  intervention  of  analogy,  agglutination  thus  creates 
syntactical  types  and  is  grammatical;  left  alone,  it  pushes  the  synthesis  of 
elements  to  the  point  where  the  elements  become  complete  units  and  produces 
only  unanalyzable  or  unproductive  words  (e.g.  hanc  horam  — >  French  encore 
'still'),  i.e.  it  is  lexicological.  [Ed.] 


absence  of  historical  evidence,  the  question  is  probably  unan- 

Only  history  can  enlighten  us.  Whenever  we  can  state  that  a 
simple  element  was  once  two  or  more  elements  in  the  sentence, 
we  have  an  agglutinate  word :  e.g.  Latin  hunc,  which  goes  back  to 
hon  ce  (ce  is  attested  epigraphically).  But  when  historical  informa- 
tion is  lacking,  it  is  hard  to  determine  what  is  due  to  agglutination 
and  what  results  from  analogy. 

Chapter  VIII 


Static  Unguistics  works  with  units  that  owe  their  existence  to  their 
sjmchronic  arrangement.  Everything  that  has  just  been  said  proves 
that  in  a  diachronic  succession  the  elements  are  not  delimited  once 
and  for  all  as  this  drawing  might  suggest : 

Period  A 

I     '^    ,  -.     Period  B 

Rather,  the  elements  are  distributed  differently  from  one  moment 
to  the  next  by  virtue  of  the  events  enacted  in  the  theatre  of  lan- 
guage, with  the  result  that  they  would  be  more  aptly  represented 
by  the  drawing: 

Period  A 

^ Period   B 

This  is  confirmed  by  all  that  has  been  said  about  the  consequences 
of  phonetic  evolution,  analogy,  agglutination,  etc. 


Almost  every  example  cited  up  to  this  point  belongs  to  word- 
formation.  Here  is  one  from  syntax.  Proto-Indo-European  had  no 
prepositions;  the  relations  that  they  indicate  were  expressed  by 
numerous  cases  that  had  great  signaling  power.  Nor  did  Proto- 
Indo-European  use  preverbs  in  compounding  verbs;  it  used  only 
particles — small  words  added  to  the  sentence  in  order  to  pinpoint 
and  modify  the  action  of  the  verb.  For  instance,  there  was  nothing 
to  correspond  to  Latin  ire  oh  inortem  'to  confront  death,'  or  to 
ohire  mortem;  the  form  would  have  been  Ire  mortem  oh.  This  was 
still  the  state  of  Proto-Greek :  (1)  In  oreos  baino  kdta,  oreos  haino 
by  itself  means  "I  come  from  the  mountain,"  the  genitive  having 
the  value  of  the  ablative;  kdta  adds  the  qualification  "by  coming 
down."  During  another  period  the  form  was  (2)  katd  oreos  haino, 
where  katd  acts  as  a  preposition,  or  even  (3)  kata-haino  oreos, 
through  the  agglutination  of  the  verb  and  particle,  which  had 
become  a  preverb. 

Here  are  found  two  or  three  distinct  phenomena,  depending  on 
the  interpretation  of  the  units:  (1)  A  new  class  of  words,  prepo- 
sitions, was  created  simply  by  shifting  existing  units.  A  particular 
arrangement  which  was  originally  of  no  significance  and  probably 
due  to  chance,  allowed  a  new  grouping:  kata,  independent  at  first, 
was  united  with  the  substantive  oreos,  and  the  whole  was  joined  to 
haino  to  serve  as  its  complement.  (2)  A  new  verbal  class  (katahaino) 
appeared.  This  is  another  psychological  grouping,  also  favored  by 
a  special  distribution  of  units  and  consolidated  by  agglutination. 
(3)  As  a  natural  consequence,  the  meaning  of  the  genitive  ending 
(6re-os)  was  weakened.  Then  katd  had  to  express  the  basic  idea 
formerly  carried  by  the  genitive  alone  and  the  importance  of  the 
ending  decreased  proportionately.  The  starting  point  of  the  future 
disappearance  of  -os  is  in  the  last  phenomenon. 

In  all  three  instances,  there  was  then  a  new  distribution  of  units. 
The  old  substance  was  given  new  functions.  The  important  thing 
is  that  no  phonetic  change  intervened  to  bring  about  any  of  the 
shifts.  But  we  must  not  think  that  meaning  alone  was  involved 
even  though  the  substance  did  not  change.  There  is  no  syntactical 
phenomenon  without  the  uniting  of  a  certain  chain  of  concepts  with 
a  certain  chain  of  phonic  units  (see  p.  139),  and  this  is  the  very 


relation  that  was  modified.  The  sounds  remained,  but  the  signi- 
ficant units  were  no  longer  the  same. 

We  saw  earlier  Cp.  75)  that  what  alters  the  sign  is  a  shift  in  the 
relationship  between  the  signifier  and  the  signified.  This  definition 
applies  not  only  to  the  alteration  of  the  terms  of  the  system  but 
also  to  the  evolution  of  the  system  itself.  The  diachronic  phe- 
nomenon in  its  totality  is  only  that  and  nothing  more. 

But  the  mere  recording  of  a  certain  shift  of  synchronic  units  is 
by  no  means  a  complete  report  of  what  has  happened  in  language. 
There  is  also  the  problem  of  the  self-contained  diachronic  unit. 
With  respect  to  every  event,  we  must  ask  which  element  has  been 
subjected  directly  to  change.  We  have  already  met  a  similar  prob- 
lem in  dealing  with  phonetic  changes  (see  p.  94).  They  affect  only 
isolated  phonemes,  leaving  the  word-unit  untouched.  Since  dia- 
chronic events  are  of  all  kinds,  many  other  such  questions  would 
have  to  be  answered,  and  the  units  delimited  in  diachrony  would 
not  necessarily  correspond  to  those  delimited  in  synchrony.  Ac- 
cording to  the  principle  laid  down  in  Part  One,  our  concept  of  the 
unit  cannot  be  the  same  in  both  cases.  In  any  event,  we  cannot  ac- 
curately define  the  unit  until  we  have  studied  it  from  both  view- 
points, the  static  and  the  evolutionary.  Until  we  solve  the  problem 
of  the  diachronic  unit,  we  cannot  penetrate  the  outer  guise  of 
evolution  and  reach  its  essence.  Understanding  units  is  just  as 
important  here  as  in  synchrony  if  we  are  to  separate  illusion  from 
reality  (see  p.  110). 

But  diachronic  identity  poses  another  difficult  question.  Indeed, 
before  I  can  say  that  a  unit  has  remained  identical  or  that  it  has 
changed  its  form  or  meaning  while  continuing  to  exist  as  a  distinct 
unit — for  both  possibilities  exist — I  must  know  the  basis  for  stating 
that  an  element  taken  from  one  period  (e.g.  French  chaud  'warm') 
is  the  same  as  an  element  taken  from  another  period  (e.g.  Latin 
calidum) . 

The  answer  will  doubtless  be  that  calidum  must  have  become 
chaud  through  regular  sound  changes  and  that  therefore  chaud  = 
calidum.  This  is  a  phonetic  identity.  The  same  applies  to  sevrer 
'wean'  and  separdre.  Fleurir  'flower,'  however,  is  not  the  same  thing 
2i,sfldrere  (which  would  have  become  *flouroir),  etc. 


Diachronic  identity  seems  at  first  glance  to  be  satisfactorily  ac- 
counted for  by  phonetic  correspondence.  But  it  is  actually  impos- 
sible for  sound  alone  to  account  for  identity.  Doubtless  it  is  correct 
to  say  that  Latin  mare  should  appear  in  French  as  mer  'sea'  because 
every  a  became  e  under  certain  conditions,  unstressed  final  e  fell, 
etc.  But  to  say  that  these  correspondences  (a  -^  e,  e  — »  zero,  etc.) 
account  for  identity  is  to  reverse  the  facts,  for  I  am  using  the 
correspondence  between  mare  and  mer  to  decide  that  a  became  e, 
that  final  e  fell,  etc. 

One  speaker  may  say  sefacher  'become  angry'  while  someone  who 
lives  in  another  part  of  France  says  se  focher,  but  this  difference  is 
unimportant  in  comparison  with  the  grammatical  facts  that  allow 
us  to  recognize  one  and  the  same  unit  of  language  in  these  two 
distinct  forms.  To  say  that  two  words  as  different  as  calidum  and 
chaud  constitute  a  diachronic  identity  means  simply  that  speakers 
passed  from  one  form  to  the  other  through  a  series  of  synchronic 
identities  in  speaking  without  there  being  a  break  in  their  common 
bond  despite  successive  phonetic  changes.  That  is  why  I  could 
state  that  knowing  how  Gentlemen!  retains  its  identity  when  re- 
peated several  times  during  a  lecture  is  just  as  interesting  as  know- 
ing why  pas  (negation)  is  identical  to  pas  (noun)  in  French,  or 
again,  why  chaud  is  identical  to  calidum  (see  p.  107  f.).  The  second 
problem  is  really  but  an  extension  and  a  complication  of  the  first. 


1.  Subjective  and  Objective  Analysis 

The  analysis  that  speakers  constantly  make  of  the  units  of  lan- 
guage is  subjective  analysis.  One  must  guard  against  confusing 
subjective  analysis  with  objective  analysis,  which  is  based  on 
history.  In  a  form  like  Greek  hippos,  the  grammarian  singles  out 
three  elements:  a  root,  a  suffix,  and  an  ending  (hipp-o-s).  But 
Greek  speakers  saw  only  two  elements  (hipp-os,  see  p.  155).  Ob- 
jective analysis  reveals  four  subunits  in  amdbds  (am-d-bd-s) ;  Latin 
speakers  recognized  only  three  (amd-bd-s) ;  perhaps  they  even 
thought  of  -bds  as  an  inflectional  whole  in  opposition  to  the  radical. 
In  French  entier  'whole'  (Latin  in-teger  'intact'),  enfant  'child' 
(Latin  in-fans  'one  who  does  not  speak'),  and  enceinte  'pregnant' 
(Latin  in-cincta  'without  a  girdle'),  the  historian  may  single  out  a 
common  prefix  en-  that  stands  for  Latin  privative  in-;  the  sub- 
jective analysis  of  speakers  completely  ignores  the  prefix. 

The  grammarian  is  prone  to  think  that  spontaneous  analyses  of 
language  are  wrong;  the  truth  is  that  subjective  analysis  is  no  more 
false  than  "false"  analogy  (see  p.  162  f.).  Language  never  errs;  it 
simply  takes  a  different  viewpoint.  There  is  no  common  yardstick 
for  both  the  analysis  of  speakers  and  the  analysis  of  the  historian 
although  both  use  the  same  procedure — the  confrontation  of  series 
that  have  a  common  element.  Both  analyses  are  justifiable,  and 
each  retains  its  value.  In  the  last  resort,  however,  only  the  speak- 
ers' analysis  matters,  for  it  is  based  directly  on  the  facts  of  lan- 

Historical  analysis  is  but  a  modified  form  of  subjective  analysis. 
Basically,  it  consists  of  projecting  the  constructions  of  different 
periods  on  a  single  plane.  It  resembles  spontaneous  analysis  in  that 
it  tries  to  identify  the  subunits  of  words  but  differs  in  that  it  syn- 
thesizes all  the  divisions  made  in  the  course  of  time  with  a  view  to 
reaching  the  oldest  one.  The  word  is  like  a  house  in  which  the 
arrangement  and  function  of  different  rooms  has  been  changed 
several  times.  Objective  analysis  adds  up  and  schematizes  the  suc- 



cessive  arrangements,  but  for  those  who  Uve  in  the  house  there  is 
always  but  one  arrangement.  The  analysis  hipp-o-s,  which  was  dis- 
cussed above,  is  not  false,  for  it  was  framed  in  the  minds  of 
speakers;  it  is  merely  "anachronistic";  it  goes  back  to  a  period 
that  preceded  the  one  from  which  the  word  is  taken.  Older  hipp-o-s 
does  not  contradict  the  hipp-os  of  Classical  Greek,  but  the  two 
analyses  cannot  be  judged  in  the  same  way.  This  again  points  up 
the  radical  distinction  between  diachrony  and  synchrony. 

And  that  allows  us  also  to  resolve  a  methodological  issue  which 
is  still  pending  in  linguistics.  The  old  school  divided  words  into 
roots,  themes,  suffixes,  etc.  and  attached  an  absolute  value  to 
these  distinctions.  One  would  think,  to  read  Bopp  and  his  disciples, 
that  the  Greeks  had  carried  with  them  from  time  immemorial  a 
collection  of  roots  and  suffixes  which  they  used  in  fabricating 
words,  and  that  they  took  the  trouble  to  manufacture  their  words 
while  speaking,  e.g.  that  pater  was  to  them  the  root  pa  +  the  suffix 
-ter,  that  doso  stood  for  the  sum  of  do  +  so  +  a  personal  end- 
ing, etc. 

There  had  to  be  a  reaction  against  the  aberrations  of  the  old 
school,  and  the  appropriate  slogan  was  this :  Observe  what  happens 
in  the  everyday  speech  of  present-day  languages  and  attribute  to 
older  periods  no  process,  no  phenomenon  that  is  not  observable 
today.  And  since  the  living  language  generally  does  not  lend  itself 
to  analyses  like  those  made  by  Bopp,  the  neogrammarians,  faithful 
to  their  principle,  declared  that  roots,  themes,  suffixes,  etc.  are 
mere  abstractions  which  should  be  used  solely  to  facilitate  ex- 
position. But  unless  there  is  some  justification  for  setting  up  these 
categories,  why  bother?  And  if  they  are  set  up,  by  what  authority 
can  one  division  like  hipp-o-s,  for  instance,  be  declared  better  than 
another  like  hipp-os? 

The  new  school,  after  pointing  out  the  shortcomings  of  the 
old  doctrine — and  this  was  easy — was  satisfied  to  reject  the  theory 
but  remain  fettered  in  practice  to  a  scientific  apparatus  that  it  was 
powerless  to  discard.  When  we  examine  "abstractions"  more 
closely,  we  see  what  part  of  reality  they  actually  stand  for,  and  a 
simple  corrective  measure  suffices  to  give  an  exact  and  justifiable 
meaning  to  the  expedients  of  the  grammarian.  That  is  what  I  have 
tried  to  do  above  by  showing  that  objective  analysis,  which  is 


intimately  linked  to  subjective  analysis  of  the  living  language,  has 
a  definite  and  rightful  place  in  linguistic  methodology. 

2.  Subjective  Analysis  and  the  Defining  of  Subunits 

In  analysis,  then,  we  can  set  up  a  method  and  formulate  defini- 
tions only  after  adopting  a  synchronic  viewpoint.  That  is  what 
I  wish  to  show  through  a  few  observations  about  word-parts: 
prefixes,  roots,  radicals,  suffixes,  and  inflectional  endings.^" 

First,  the  inflectional  ending,  i.e.  the  word-final  variable  element 
that  distinguishes  the  different  forms  of  a  noun  or  verb  paradigm. 
In  zeugnu-mi,  zeugnu-s,  zeugnu-si,  zeugnu-men,  etc.  'I  harness,'  etc., 
the  inflectional  endings  -mi,  -s,  -si,  etc.  stand  out  simply  because 
they  are  in  opposition  to  each  other  and  to  the  preceding  part  of 
the  word  (zevgnu-).  We  recall  that  in  Czech  the  absence  of  an  in- 
flectional ending  plays  the  same  role  as  a  regular  ending  (e.g.  the 
genitive  plural  zen  in  opposition  to  nominative  singular  zena;  see 
p.  86  and  p.  118).  Similarly,  Greek  zeugnU!  '(thou)  harness!' 
against  zeitgnu-te  '(you)  harness!'  or  rhetor!  against  rhetor-os,  etc. 
and  French  marl!,  written  marche  '(thou)  walk!'  against  mar  so! 
'(let's)  walk!'  are  all  inflected  forms  with  a  zero  ending. 

By  eliminating  the  inflectional  ending  we  obtain  the  inflectional 
theme  or  radical.  This  is  generally  the  common  element  which 
emerges  spontaneously  when  we  compare  a  series  of  related  words, 
whether  inflected  or  not,  and  which  conveys  the  idea  common  to 
every  word.  In  the  French  series  roulis  'roll,'  rouleau  'rolling-pin,' 
r outage  'roller,'  roulement  'rolling,'  for  instance,  the  radical  roul- 
stands  out.  But  in  their  analysis,  speakers  often  single  out  several 
kinds,  or  rather  grades,  of  radicals  in  the  same  family  of  words. 
Zeugnu-,  separated  above  from  zeugnu-mi,  zeugnu-s,  etc.,  is  a  first- 
grade  radical.  It  is  not  irreducible,  for  the  division  zeug-nu  is  self- 
evident  if  we  compare  zeugnu-  with  other  series  {zeugnumi,  zeuk- 
tos,  zeuksis,  zeukter,  zugon,  etc.  on  the  one  hand  and  zeugnUmi, 

1"  F.  de  Saussure  did  not  study  the  question  of  compounds — not  from  the 
synchronic  viewpoint  at  any  rate.  This  part  of  the  problem  must  therefore 
be  set  aside.  Of  course  the  distinction  made  above  between  compounds  and 
agglutinate  words  does  not  apply  here  where  analysis  of  a  language-state  is 
concerned.  It  is  scarcely  necessary  to  point  out  that  this  account  of  subunits 
does  not  pretend  to  answer  the  more  difficult  question  raised  above  (pp.  105, 
110  f.)  concerning  the  defining  of  the  word-unit.  [FA.] 


deiknumi,  ornumi,  etc.  on  the  other).  Zeug-  (with  its  alternate 
forms  zeug-,  zeuk-,  zug-;  see  p.  160)  is  therefore  a  second-grade 
radical.  But  zeug-  is  irreducible.  To  carry  its  decomposition  further 
by  comparing  related  forms  is  not  possible. 

The  root  is  the  irreducible  element  common  to  all  words  of  the 
same  family.  But  any  subjective  and  synchronic  analysis  separates 
material  elements  only  by  considering  the  share  of  meaning  that 
matches  each  element,  and  the  root  is  in  this  respect  the  element 
in  which  the  meaning  common  to  all  related  words  reaches  the 
highest  degree  of  abstraction  and  generality.  Naturally,  indefinite- 
ness  varies  from  one  root  to  the  next,  but  it  also  depends  somewhat 
on  the  extent  to  which  the  radical  is  reducible.  The  more  the  radical 
is  shortened,  the  greater  the  likehhood  that  its  meaning  will  become 
abstract.  Thus  zeugmdtion  suggests  a  little  team,  zeugma  any  team 
whatsoever,  and  zeug-  the  indefinite  notion  of  yoking  or  harnessing. 

It  follows  that  a  root  cannot  constitute  a  word  and  have  an 
inflectional  ending  joined  directly  to  it.  Indeed,  a  word  always 
stands  for  a  fairly  definite  idea,  at  least  from  a  grammatical  view- 
point, and  this  is  contrary  to  the  general  and  abstract  nature  of  the 
root.  But  what  about  the  numerous  roots  and  inflectional  themes 
that  apparently  mingle?  Take  Greek  phloks,  genitive  phlogos 
against  the  root  phleg-:  phlog-  which  is  found  in  every  word  of  the 
same  family  (cf.  phleg-o,  etc.).  Does  this  not  contradict  the  dis- 
tinction which  we  have  just  set  up?  No,  for  we  must  separate 
phleg-:  phlog-  with  a  general  meaning  from  phlog-  with  its  special 
meaning  or  risk  considering  the  material  form  only  to  the  exclusion 
of  meaning.  The  same  material  element  here  has  two  different 
values.  It  therefore  comprises  two  distinct  linguistic  elements 
(see  p.  105).  Above,  it  was  shown  that  zeugnu!  is  a  word  with  an 
inflectional  ending  of  zero.  In  the  same  way,  phlog-  is  a  theme  with 
a  zero  suffix.  No  confusion  is  possible.  The  radical  is  distinct  from 
the  root  even  when  phonetically  identical  to  it. 

The  root  is  then  a  reality  in  the  mind  of  speakers.  To  be  sure, 
speakers  do  not  always  single  it  out  with  equal  precision.  On  this 
point  there  are  differences,  either  within  the  same  language  or 
from  one  language  to  another. 

In  certain  idioms,  definite  characteristics  call  the  root  to  the 
attention  of  speakers.  In  German,  for  instance,  the  root  is  fairly 


uniform;  almost  always  monosyllabic  (cf.  streit-,  hind-,  haft-,  etc.), 
it  follows  certain  structural  rules;  phonemes  do  not  appear  hap- 
hazardly; certain  word-final  combinations  of  consonants,  such  as 
occlusive  -f-  liquid,  are  ruled  out;  werk-  is  possible,  wekr-  is  not; 
we  find  helf-,  werd-,  but  not  hefl-,  wedr-. 

We  recall  that  regular  alternations,  especially  between  vowels, 
tend  generally  to  strengthen  rather  than  to  weaken  our  feeling  for 
roots  and  subunits.  Here  also,  German  with  its  variable  interplay 
of  ablauts  (see  p.  158)  differs  greatly  from  French.  Semitic  roots 
exhibit  the  same  characteristic  but  in  even  greater  proportions. 
Here  the  alternations  are  quite  regular  and  govern  a  large  number 
of  complex  oppositions  (cf.  Hebrew  qdtal,  qtaltem,  qtol,  qitlu,  etc., 
all  forms  of  the  same  verb  meaning  'kill').  In  addition,  Semitic 
roots  have  a  trait  similar  to  German  monosyllabism  but  even  more 
striking.  They  always  include  three  consonants  (see  below, 
pp.  230  ff.). 

French  is  completely  different.  It  has  few  alternations  and,  side 
by  side  with  monosyllabic  roots  (roul-,  march-,  mang-),  many  roots 
composed  of  two  or  even  three  syllables  {commenc-,  hesit-,  epou- 
vant-).  Besides,  these  roots  contain — chiefly  in  final  position — such 
varied  combinations  that  they  cannot  be  reduced  to  rules  (cf .  tu-er 
'kill,'  regn-er  'reign,'  guid-er  'guide,'  grond-er  'growl,'  souffl-er 
'blow,'  tard-er  'delay,'  entr-er  'enter,'  hurl-er  'bark,'  etc.).  That  the 
feeling  for  roots  scarcely  exists  in  French  should  come  as  no 

The  defining  of  the  root  has  as  its  counterpart  the  defining 
of  prefixes  and  suffixes.  The  prefix  goes  before  the  part  of  the 
word  that  is  recognized  as  the  radical  (e.g.  hupo-  in  Greek  hupo- 
zeugnwni).  The  suffix  is  the  element  added  to  the  root  to  make  a 
radical  (e.g.  zeug-mat-)  or  to  a  first-grade  radical  to  make  a  second- 
grade  radical  (e.g.  zeugmat-io-) .  We  saw  above  that  the  suffix,  like 
the  inflectional  ending,  may  be  zero.  The  extracting  of  the  suffix  is 
just  one  more  side  to  the  analysis  of  the  radical. 

The  suSix  sometimes  has  a  concrete  meaning,  a  semantic  value, 
as  in  zeuk-ter,  where  -ter-  names  the  agent  or  performer  of  an  ac- 
tion. At  other  times  the  suffix  has  a  mere  grammatical  function,  as 
in  zeug-nu  (-^mi),  where  -nu  expresses  the  idea  of  the  present.  The 
prefix  may  also  play  both  roles,  but  our  languages  rarely  give  it  a 


grammatical  function:  e.g.  the  ge-  of  German  past  participles 
{ge-setzt,  etc.),  the  perfective  prefixes  of  Slavic  (Russian  na-pisdt', 

The  prefix  also  differs  from  the  suffix  through  a  characteristic 
which,  though  fairly  general,  is  not  absolute.  The  prefix  is  more 
sharply  delimited,  for  it  is  easier  to  separate  from  the  word  as  a 
whole.  This  is  due  to  the  very  nature  of  the  prefix.  A  complete  word 
usually  remains  after  the  prefix  is  removed  (cf .  French  recommencer 
'recommence':  commencer  'commence,'  indigne  'unworthy':  digne 
'worthy,'  maladroit  'unskilled' :  adroit  'skilled,'  contrepoids  'counter- 
weight': poids  'weight,'  etc.).  Latin,  Greek,  and  German  offer  even 
more  striking  examples.  Moreover,  many  prefixes  function  as  inde- 
pendent words:  cf.  French  centre  'against,'  mal  'ill,'  avant  'before,' 
sur  'on,'  German  unter,  vor,  etc.,  and  Greek  katd,  pro,  etc.  But  the 
suffix  is  altogether  different.  The  radical  element  obtained  by  re- 
moving the  suffix  is  not  a  complete  word :  e.g.  French  organisation 
'organization' :  organis-,  German  Trennung:  trenn-,  Greek  zeugma: 
zeug-,  etc.^^  Furthermore,  the  suffix  has  no  independent  existence. 

The  result  is  that  the  first  part  of  the  radical  is  usually  delimited 
beforehand.  The  speaker  knows,  before  he  has  made  any  com- 
parisons with  other  forms,  where  to  draw  the  line  between  the  pre- 
fix and  what  follows.  This  is  not  true  of  the  last  part  of  the  word. 
There  one  can  draw  no  boundary  without  first  comparing  forms 
that  have  the  same  radical  or  suffix,  and  the  resulting  delimitations 
will  vary  according  to  the  nature  of  the  terms  compared. 

Subjectively,  suffixes  and  radicals  derive  their  value  solely  from 
syntagmatic  and  associative  oppositions.  We  can  usually  /ind  a 
formative  and  a  radical  element  in  any  two  opposing  parts  of  a 
word,  provided  that  possible  oppositions  exist.  In  Latin  diddtorem, 
for  instance,  we  shall  see  the  radical  dictdtdr-{em)  if  we  compare  it 
with  consul-em,  ped-em,  etc.;  dicta-{tdrem)  if  we  compare  it  with 
lic-torem,  scrip-torem,  etc.;  and  dic-{tdtdrem)  if  we  think  of  po- 
tdtorem,  can-tdiorem,  etc.  Generally,  and  under  favorable  circum- 
stances, the  speaker  may  make  every  imaginable  division  (e.g. 
dictdt-orem,  from  am-orem,  ard-drem,  etc.;  dict-dtdrem,  from  dr- 

"  This  pattern,  though  not  necessarily  applicable  to  English  words  derived 
from  Germanic  sources  (teach-er,  sad-ly,  hope-less),  is  characteristic  of  English 
words  derived  from  Romance  sources  {duch-ess,  appari-tion,  cap-able).  [Tr.] 


dtdrem,  ar-dtorem,  etc.)-  We  know  that  the  results  of  these  sponta- 
neous analyses  appear  in  the  analogical  formations  of  each  period 
(see  p.  170).  Through  them,  we  can  single  out  the  subunits  (roots, 
prefixes,  suffixes,  and  endings)  which  language  recognizes  and  the 
values  which  it  attaches  to  them. 

3.  Etymology 

Etymology  is  neither  a  distinct  discipline  nor  a  division  of  evolu- 
tionary linguistics.  It  is  only  a  special  application  of  the  principles 
that  relate  to  synchronic  and  diachronic  facts.  It  goes  back  into 
the  history  of  words  until  it  finds  something  to  explain  them. 

To  speak  of  the  origin  of  a  word  and  say  that  it  "comes"  from 
another  word  may  imply  several  different  things :  thus  French  sel 
comes  from  Latin  sal  through  a  simple  sound  change;  labourer 
'plough'  comes  from  Old  French  labourer  'work'  solely  through  a 
change  in  meaning;  couver  'brood'  comes  from  Latin  cubare  'be  in 
bed'  through  a  change  in  both  meaning  and  sound;  finally,  the 
statement  that  French  pommier  'apple-tree'  comes  from  pomme 
'apple'  brings  in  the  relation  of  grammatical  derivation.  The  first 
three  examples  concern  diachronic  identities;  the  fourth  is  based 
on  the  synchronic  relation  of  several  different  terms,  and  every- 
thing that  has  been  said  about  analogy  shows  that  this  relation  is 
the  most  important  part  of  et3Tnological  research. 

It  is  not  possible  to  fix  the  etymology  of  bonus  merely  by  going 
back  to  dvenos.  But  if  bis  is  found  to  go  back  to  dvis,  implying  a 
relation  with  duo,  then  the  procedure  is  etymological.  The  same 
applies  to  the  comparing  of  French  oiseau  'bird'  and  Latin  avi- 
cellus,  for  comparison  reveals  the  link  between  oiseau  and  avis. 

Etjnnology  is  then  mainly  the  explaining  of  words  through  the 
historical  study  of  their  relations  with  other  words.  To  explain 
means  to  relate  to  known  terms,  and  in  linguistics,  to  explain  a  word 
is  to  relate  it  to  other  words,  for  there  are  no  necessary  relations 
between  sound  and  meaning  (principle  of  the  arbitrary  nature  of 
the  sign,  see  p.  67  f.). 

Etymology  does  not  simply  explain  isolated  words  and  stop 
there.  It  compiles  the  history  of  word  families  and  of  families  of 
formative  elements — prefixes,  suffixes,  etc. 

Like  static  and  evolutionary  linguistics,  etymology  describes 


facts.  But  this  description  is  not  methodical,  for  it  follows  no  fixed 
course.  In  compiling  the  history  of  a  word,  etymology  borrows  its 
data  alternately  from  phonetics,  morphology,  semantics,  etc.  To 
reach  its  goal,  etjnnology  uses  every  means  placed  at  its  disposal 
by  linguistics,  but  it  is  not  concerned  with  the  nature  of  the 
operations  that  it  is  obliged  to  perform. 

Geographical  Linguistics 

Chapter  I 

As  we  approach  the  question  of  the  spatial  relations  of  the  lin- 
guistic phenomenon,  we  leave  internal  linguistics  and  enter  ex- 
ternal linguistics.  The  scope  of  external  linguistics  was  outlined 
in  Chapter  V  of  the  Introduction. 

The  most  striking  thing  about  the  study  of  languages  is  their 
diversity — linguistic  differences  that  appear  when  we  pass  from  one 
country  to  another  or  even  from  one  region  to  another.  Divergences 
in  time  often  escape  the  observer,  but  divergences  in  space  im- 
mediately force  themselves  upon  him;  even  savages  grasp  them, 
thanks  to  their  contacts  with  other  tribes  that  speak  a  different 
language.  Indeed,  these  comparisons  are  what  makes  a  nation 
aware  of  its  idiom. 

We  note  in  passing  that  this  feeling  makes  primitive  people  look 
upon  language  as  a  habit  or  custom  like  dress  or  weapons.  The  term 
idiom  rightly  designates  language  as  reflecting  the  traits  peculiar 
to  a  community  (Greek  idioma  had  already  acquired  the  meaning 
'special  custom').  This  notion,  though  appropriate,  becomes  mis- 
leading when  one  goes  so  far  as  to  see  language  as  an  attribute,  not 
of  the  nation,  but  of  race,  in  the  same  way  as  the  color  of  the  skin 
or  the  shape  of  the  head. 

It  is  also  worth  noting  that  each  nation  believes  in  the  su- 
periority of  its  own  idiom  and  is  quick  to  regard  the  man  who  uses 
a  different  language  as  incapable  of  speaking.  For  instance,  Greek 
hdrbaros  apparently  meant  'one  who  stammers'  and  was  related  to 
Latin  balbus;  in  Russian,  Germans  are  called  Nemtsy  'mutes.' 

Geographical  diversity  was,  then,  the  first  observation  made  in 
linguistics.  It  determined  the  initial  form  of  scientific  research  in 
language,  even  among  the  Greeks.  To  be  sure,  the  Greeks  were 



concerned  only  with  the  diversity  of  the  different  Hellenic  dialects, 
but  this  was  because  their  interest  did  not  generally  go  beyond  the 
borders  of  Greece  proper. 

Having  noticed  that  two  idioms  differ,  one  instinctively  looks 
for  similarities.  This  is  a  natural  tendency  of  speakers.  Peasants 
Uke  to  compare  their  patois  with  the  one  spoken  in  a  neighboring 
village.  People  who  speak  several  languages  notice  their  common 
traits.  But  for  some  strange  reason  science  has  waited  a  long  time 
to  make  use  of  the  results  of  such  observations.  For  example,  the 
Greeks  noticed  many  resemblances  between  the  Latin  vocabulary 
and  their  own  but  were  unable  to  draw  any  linguistic  conclusions. 

Scientific  observation  of  linguistic  similarities  proves  that  two 
or  more  idioms  may  be  akin,  i.e.  that  they  have  a  common  origin. 
A  group  of  related  languages  makes  up  a  family.  Modern  linguistics 
has  successively  identified  several  families:  the  Indo-European, 
Semitic,  Bantu, ^  etc.  Comparing  these  families  with  each  other,  in 
turn,  occasionally  brings  to  light  older  and  broader  affiliations. 
There  have  been  attempts  to  find  similarities  between  Finno-Ugric^ 
and  Indo-European,  between  the  latter  and  Semitic,  etc.,  but  such 
comparisons  always  come  up  against  insuperable  barriers.  One 
must  not  confuse  what  is  probable  with  what  is  demonstrable.  The 
universal  kinship  of  languages  is  not  probable,  but  even  if  it  were 
true — as  the  Italian  linguist  Trombetti^  believes — it  could  not  be 
proved  because  of  the  excessive  number  of  changes  that  have 

Beside  diversity  within  related  groups,  then,  there  is  absolute 
diversity — differences  between  languages  that  have  no  recognizable 
or  demonstrable  kinship.  What  method  should  linguistics  use  in 
each  of  these  degrees?  Let  us  begin  with  the  second,  which  is  more 
common.  As  we  have  just  noted,  countless  languages  and  families  of 

^  Bantu  is  a  group  of  languages  spoken  by  South  African  tribes,  mainly  the 
Kaffirs.  [Ed.] 

"^  Finno-Ugric,  which  includes — among  other  languages — Finnish  proper  or 
Suomi,  Mordvinian,  Lapp,  etc.,  is  a  family  of  languages  spoken  in  northern 
Russia  and  Siberia.  Doubtless  these  languages  all  go  back  to  a  common 
original  idiom.  The  family  is  a  part  of  the  great  Ural-Altaic  group  of  languages, 
which  have  no  proven  common  origin  although  some  traits  appear  in  all  of 
them.  [Ed.] 

'  See  his  L'unitd,  d'origine  del  linguaggio,  Bologna,  1905.  [Ed.] 


languages  are  not  related.  A  good  example  is  Chinese  with  respect 
to  the  Indo-European  languages.  The  fact  that  they  differ  does  not 
mean  that  they  cannot  be  compared,  for  comparison  is  always  pos- 
sible and  useful;  it  applies  to  grammatical  organisms  and  general 
ways  of  expressing  thought  as  well  as  to  systems  of  sound ;  it  also 
includes  diachronic  facts,  the  phonetic  evolution  of  two  languages, 
etc.  The  possibilities  of  comparison,  though  incalculable,  are 
limited  by  certain  constant  phonic  and  psychological  data  that 
determine  the  make-up  of  any  language ;  reciprocally,  the  discovery 
of  these  constant  data  is  always  the  main  aim  of  any  comparison  of 
related  languages. 

The  other  class  of  differences — those  that  exist  within  families  of 
languages — offers  an  unlimited  field  for  comparison.  Two  idioms 
may  differ  in  any  degree.  They  may  bear  a  striking  resemblance  to 
each  other,  like  Zend  and  Sanskrit,  or  be  as  entirely  dissimilar  as 
Sanskrit  and  Gaelic.  All  intermediate  degrees  are  possible:  Greek 
and  Latin  are  more  closely  related  to  each  other  than  to  Sanskrit, 
etc.  Idioms  that  differ  only  slightly  are  called  dialects,  but  this 
word  must  be  used  loosely.  We  shall  see  that  languages  and  dialects 
differ  quantitatively,  not  by  nature  (see  p.  203). 

Chapter  II 


1.  Coexistence  of  Several  Languages  at  the  Same  Point 

Up  to  this  point  geographical  diversity  has  been  presented  in  its 
ideal  form:  there  were  as  many  territories  as  there  were  different 
languages.  And  our  method  was  justifiable,  for  geographical  sepa- 
ration is  still  the  most  general  force  in  linguistic  diversity.  But 
there  are  secondary  facts  that  disturb  the  ideal  relationship  and 
cause  several  languages  to  coexist  in  the  same  territory. 

Two  things  we  pass  over.  First  is  the  real,  organic  mixture  or 
interpenetration  of  two  idioms  that  results  in  a  change  in  the 


system  (cf.  English  after  the  Norman  Conquest).  Second  is  the 
political  accident  of  several  languages  clearly  separated  in  space 
but  included  within  the  boundaries  of  the  same  state,  as  in  Switzer- 
land. The  only  fact  that  concerns  us  is  that  two  idioms  can  exist 
side  by  side  in  the  same  place  without  intermingling.  This  occurs 
frequently,  but  is  of  two  kinds. 

First,  newcomers  may  superimpose  their  language  on  the  indig- 
enous language.  For  instance,  in  South  Africa,  two  successive 
colonizations  introduced  Dutch  and  English,  which  now  exist 
alongside  several  Negro  dialects;  in  the  same  way,  Spanish  was 
implanted  in  Mexico.  Nor  are  such  linguistic  encroachments  pe- 
cuhar  to  modern  times.  Throughout  the  centuries  nations  have 
intermingled  and  still  kept  their  idioms  distinct.  To  realize  this  fact 
we  need  only  glance  at  a  map  of  modern  Europe:  Ireland,  with 
Celtic  and  English;  many  of  the  Irish  speak  both  languages.  In 
Brittany,  French  and  Breton.  In  the  Basque  region,  French  and 
Spanish  as  well  as  Basque.  In  Finland,  Swedish  and  Finnish  have 
coexisted  for  a  rather  long  time,  and  Russian  has  been  added  more 
recently.  In  Courland  and  Livonia,  Lettish,  German  and  Russian 
are  spoken;  German,  which  was  brought  in  by  colonists  under  the 
auspices  of  the  Hanseatic  League  during  the  Middle  Ages,  belongs 
to  a  special  segment  of  the  population;  Russian  subsequently 
entered  by  conquest.  Lithuania  witnessed  the  implantation  of 
Polish  alongside  Lithuanian  as  a  consequence  of  her  former  union 
with  Poland,  and  of  Russian  as  a  result  of  annexation.  Until  the 
eighteenth  century  Slavic  and  German  were  used  throughout  the 
section  of  Germany  that  lies  to  the  east  of  the  Elbe.  In  other 
countries  languages  are  even  more  entangled:  in  Macedonia  every 
imaginable  language  is  found — Turkish,  Bulgarian,  Serbian,  Greek, 
Albanian,  Rumanian,  etc. — and  the  languages  are  mixed  in 
different  ways  in  different  regions. 

Coexisting  languages  are  not  always  absolutely  entangled;  there 
may  be  a  certain  relative  territorial  distribution.  Of  two  languages, 
one  may  be  spoken  in  town  and  the  other  in  the  country,  but  such 
a  distribution  is  not  always  clear-cut. 

The  story  was  the  same  in  ancient  times.  A  linguistic  map  of  the 
Roman  Empire  would  show  facts  like  those  already  described. 
Toward  the  close  of  the  Republic,  for  instance,  Campania  num- 


bered  three  or  four  languages:  Oscan,  attested  by  the  inscriptions 
of  Pompeii;  Greek,  the  language  of  the  colonists  who  founded 
Naples,  etc.;  Latin;  and  perhaps  even  Etruscan,  which  was  the 
dominant  language  before  the  arrival  of  the  Romans.  In  Carthage, 
Punic  or  Phoenician  persisted  beside  Latin  (it  still  existed  during 
the  period  of  the  Arab  invasion),  and  Numidian  was  certainly- 
spoken  in  Carthaginian  territory.  One  might  also  suppose  that 
during  ancient  times  unilingual  countries  in  the  Mediterranean 
Basin  were  the  exception. 

Invasion  is  the  usual  cause  of  superimposition,  but  it  may  also 
come  through  peaceful  penetration  in  the  form  of  colonization.  Or 
nomadic  tribes  may  take  their  dialect  with  them :  that  is  what  the 
Gypsies  did,  especially  those  who  settled  in  Hungary,  where  they 
form  compact  villages;  study  of  their  language  shows  that  they 
must  have  come  from  India  at  some  unknown  time  in  the  past.  In 
Dobruja,  at  the  mouth  of  the  Danube,  scattered  Tatar  villages 
show  up  Uke  tiny  specks  on  the  hnguistic  map  of  the  region. 

2.  Literary  Language  and  Local  Idiom 

As  a  further  step,  linguistic  unity  may  be  destroyed  when  a 
natural  idiom  is  influenced  by  a  literary  language.  This  never  fails 
to  happen  whenever  a  nation  reaches  a  certain  stage  of  civilization. 
By  literary  language  I  mean  not  only  the  language  of  literature  but 
also,  in  a  more  general  sense,  any  kind  of  cultivated  language, 
official  or  otherwise,  that  serves  the  whole  community.  Given  free 
reign,  a  language  has  only  dialects,  none  of  which  has  the  advan- 
tage over  the  others,  and  for  this  reason  it  habitually  splinters.  But 
as  communications  improve  with  a  growing  civiUzation,  one  of  the 
existing  dialects  is  chosen  by  a  tacit  convention  of  some  sort  to  be 
the  vehicle  of  everything  that  affects  the  nation  as  a  whole.  The 
reasons  for  the  choice  differ  widely.  Sometimes  preference  goes  to 
the  dialect  of  the  region  where  civilization  is  most  advanced  or  to 
the  province  that  has  political  supremacy  and  wields  the  central 
power.  Sometimes  the  court  imposes  its  dialect  on  the  nation.  The 
privileged  dialect,  after  it  has  been  promoted  to  the  rank  of  official 
and  standard  language,  seldom  remains  the  same  as  it  was  before. 
It  acquires  dialectal  elements  from  other  regions  and  becomes  more 
and  more  composite,  though  without  losing  completely  its  original 


character.  Thus  the  dialect  of  the  He  de  France  is  clearly  recogniz- 
able in  literary  French  and  the  Toscan  in  Standard  Italian.  But  the 
literary  language  is  not  imposed  from  one  day  to  the  next,  and  a 
majority  of  the  population  is  found  to  be  bilingual,  speaking  both 
the  standard  language  and  the  local  patois.  This  occurs  in  many 
parts  of  France,  like  Savoy,  where  French  is  an  imported  language 
that  has  not  yet  eliminated  the  regional  patois,  and  generally  in 
Germany  and  Italy,  where  dialects  persist  alongside  the  official 

It  has  been  the  same  with  all  nations  that  have  reached  a  certain 
stage  of  civihzation.  The  Greeks  had  their  koine,  derived  from  Attic 
and  Ionian,  along  with  coexisting  local  dialects.  Presumably  even 
ancient  Babylon  had  its  official  language  and  its  regional  dialects. 

Does  a  standard  language  necessarily  imply  the  use  of  writing? 
The  Homeric  poems  seem  to  prove  that  it  does  not.  Even  though 
they  were  composed  at  a  time  when  writing  was  used  little  or  not 
at  all,  their  language  is  conventional  and  has  every  characteristic 
of  a  literary  language. 

The  facts  discussed  in  this  chapter  are  so  common  that  they 
might  pass  as  normal  forces  in  the  history  of  languages.  But  to  keep 
to  our  purpose  we  must  turn  aside  from  everything  that  obscures 
the  basic  phenomenon  of  natural  geographical  diversity  and  con- 
sider it  apart  from  any  importation  of  a  foreign  language  or  any 
formation  of  a  literary  language.  This  schematic  simplification 
seems  to  go  against  reality,  but  the  natural  fact  must  first  be 
studied  in  itself. 

Consistently  with  this  principle,  we  shall  say  that  Brussels  is 
Germanic  since  it  is  in  the  Flemish  part  of  Belgium ;  though  French 
is  spoken  there,  what  matters  is  the  boundary  between  the  Flemish 
and  Walloon  territories.  Li^ge  is  Romance  for  the  same  reason :  it  is 
in  Walloon  territory ;  French  is  a  foreign  language  that  happens  to 
have  been  superimposed  on  a  dialect  of  the  same  stock.  Similarly, 
Brest  belongs  linguistically  to  Breton ;  the  French  spoken  there  has 
nothing  in  common  with  the  native  idiom  of  Brittany.  Berlin, 
where  High  German  is  heard  almost  exclusively,  is  Low  German, 



Chapter  III 

1.  Time,  the  Basic  Cause 

Whereas  absolute  diversity  poses  a  purely  speculative  problem 
(see  p.  192  f.),  diversity  within  related  languages  can  be  observed 
and  traced  back  to  unity.  That  Vulgar  Latin  took  different  paths 
in  the  northern  and  southern  parts  of  Gaul  explains  the  common 
origin  of  French  and  Proven  gal. 

By  simplifying  the  theoretical  situation  as  much  as  possible,  we 
can  get  at  the  basic  cause  of  differentiation  in  space.  What  would 
happen  if  a  language  spoken  at  one  clearly  delimited  point — e.g.  a 
small  island — were  transported  by  colonists  to  another  clearly  de- 
limited point — e.g.  another  island?  After  a  certain  length  of  time 
various  differences  affecting  vocabulary,  grammar,  pronunciation 
and  the  like  would  separate  the  language  of  the  source  (S)  from  the 
language  of  the  settlement  (S'). 

It  is  wrong  to  imagine  that  only  the  transplanted  idiom  will 
change  while  the  original  idiom  remains  fixed  or  vice  versa.  An 
innovation  may  begin  on  either  side  or  on  both  sides  at  the  same 
time.  Take  a  linguistic  feature  a  that  can  be  replaced  by  h,  c,  d,  etc. 
Differentiation  may  occur  in  three  different  ways : 

a  (Source  S) 
o  (Settlement  S') 

A  one-sided  approach  will  not  do,  for  the  innovations  of  either 
language  are  of  equal  importance. 

What  created  the  differences?  It  is  illusory  to  think  that  space 


alone  was  responsible.  By  itself,  space  cannot  influence  language*. 
On  the  day  following  their  arrival  at  S'  the  colonists  from  S  spoke 
exactly  the  same  language  as  on  the  preceding  day.  It  is  easy  to 
forget  about  the  factor  of  time  because  it  is  less  concrete  than 
space,  but  it  is  actually  the  cause  of  linguistic  differentiation. 
Geographical  diversity  should  be  called  temporal  diversity. 

Take  two  differentiating  features  h  and  c.  No  speakers  have 
passed  from  the  first  to  the  second  or  from  the  second  to  the  first. 
To  discover  how  unity  became  diversity,  we  must  go  back  to  the 
original  a  for  which  b  and  c  were  substituted:  a  gave  way  to  the 
later  forms  b  and  c.  Hence  the  following  diagram  of  geographical 
differentiation  which  will  cover  all  similar  cases : 

S  S' 

a< — >  a 

i  I 

b  c 

The  separation  of  the  two  idioms  shows  the  tangible  form  of  the 
phenomenon  but  does  not  explain  it.  Undoubtedly  divergence  in 
space  was  a  necessary  condition — no  matter  how  small  the  amount 
— but  by  itseff  distance  does  not  create  differences.  Volume  is 
measured,  not  by  one  surface,  but  by  adding  a  third  dimension, 
depth,  similarly,  geographical  differentiation  is  pictured  com- 
pletely only  when  projected  in  time. 

One  objection  might  be  that  differences  in  environment,  climate, 
topography,  and  local  customs  (e.g.  customs  of  mountaineers  con- 
trasted with  those  of  a  maritime  population)  influence  language, 
and  that  our  variations  are  therefore  conditioned  geographically. 
Such  influences  are  open  to  dispute,  however  (see  p.  147  f.).  Even  if 
they  could  be  proved,  a  further  distinction  would  be  in  order: 
direction  of  movement,  which  is  governed  in  each  instance  by  im- 
ponderable forces  that  can  neither  be  demonstrated  nor  described, 
is  attributable  to  environment.  At  a  particular  moment  and  in  a 
particular  environment  u  became  it.  Why  did  it  change  at  that 
moment  and  in  that  place,  and  why  did  it  become  ii  instead  of  of 
That  question  we  cannot  answer.  But  change  itself  (leaving  out  the 
special  direction  it  takes  and  its  particular  manifestations) — in 



short,  the  instability  of  language — stems  from  time  alone.  Geo- 
graphical diversity  is  then  a  secondary  side  of  the  general  phe- 
nomenon. The  unity  of  related  languages  is  found  only  in  time. 
Unless  the  comparative  linguist  thoroughly  assimilates  this  princi- 
ple, he  is  likely  to  delude  himself. 

2.  Effect  of  Time  on  Continuous  Territory 

Now  take  a  unilingual  country,  i.e.  one  with  a  uniform  language 
and  a  stable  population,  hke  Gaul  around  450  a.d.,  when  Latin 
was  well  established  everywhere.  What  will  happen? 

(1)  Since  there  is  no  such  thing  as  absolute  immobility  in  speech 
(see  pp.  75  ff.),  the  language  will  no  longer  be  the  same  after  a 
certain  length  of  time. 

(2)  Evolution  will  not  be  uniform  throughout  the  territory  but 
will  vary  from  zone  to  zone ;  no  records  indicate  that  any  language 
has  ever  changed  in  the  same  way  throughout  its  territory.  There- 
fore, it  is  not  the  diagram : 

but  the  diagram : 

that  gives  the  true  picture. 

How  do  differences  that  result  in  the  most  varied  dialectal  forms 
originate?  What  pattern  does  their  evolution  follow?  Differentia- 
tion through  time,  which  is  not  so  simple  as  it  seems  at  first,  has 
two  main  characteristics: 



(1)  Evolution  takes  the  form  of  successive  and  precise  inno- 
vations that  include  as  many  partial  facts  as  could  be  enumerated, 
described,  and  classified  according  to  their  nature  (phonetic,  lexico- 
logical, morphological,  syntactical,  etc.). 

(2)  Each  innovation  embraces  a  definite  and  delimited  area. 
There  are  two  possibilities:  either  the  area  of  the  innovation  em- 
braces the  whole  territory  and  creates  no  dialectal  differences  (the 
less  usual  possibility) ,  or  the  change  affects  only  a  part  of  the  ter- 
ritory, each  dialectal  fact  having  its  special  zone  (the  more  common 
occurrence).  We  can  illustrate  with  phonetic  changes,  but  other 
innovations  are  the  same.  For  instance,  while  part  of  a  territory 
may  witness  the  change  of  a  to  e: 

it  is  possible  that  on  the  same  territory  but  within  other  limits, 
another  change,  such  as  s  to  z;,  will  occur: 

and  the  existence  of  these  distinct  areas  explains  the  diversity  of 
regional  speech-forms  throughout  the  territory  of  a  language  that 
is  allowed  to  evolve  naturally.  There  is  no  way  to  foresee  these 
zones;  nothing  points  to  which  way  they  will  spread;  all  we  can  do 
is  record  them.  Laid  on  a  map,  with  their  boundaries  crossing  and 
recrossing  each  other,  they  form  extremely  complicated  patterns. 
At  times  their  configuration  is  paradoxical.  Thus  c  and  g  changed 
before  a  to  th,  dz,  then  h,  z  (cf.  cantum  —^  chant  'song,'  virga  — »  verge 
'rod')  throughout  northern  France  except  in  Picardy  and  part  of 
Normandy,  where  c  and  g  remained  intact  (cf .  Picard  cat  for  chat 


'cat,'  rescape  for  rechappe,  which  was  recently  adopted  by  French,* 
vergue  from  virga,  cited  above,  etc.)- 

What  is  the  result  of  differentiation  through  time?  At  one 
moment  in  history  a  single  language  may  reign  throughout  a 
particular  territory,  and  five  or  ten  centuries  later  the  inhabitants 
of  two  of  its  extremes  probably  vnW  not  be  able  to  understand  each 
other.  At  any  particular  point,  however,  speakers  will  still  under- 
stand the  speech-forms  of  neighboring  regions.  A  traveler  going 
from  one  end  of  the  country  to  the  other  would  notice  only  small 
dialectal  differences  from  one  locality  to  the  next.  But  the  sum  of 
these  differences  would  increase,  and  eventually  he  would  come  to 
a  language  that  the  inhabitants  of  this  starting  point  would  not 
understand.  Or  if,  starting  from  a  given  point  in  the  territory,  he 
traveled  outward,  now  in  one  direction,  now  in  another,  he  would 
find  the  sum  of  these  differences  increasing  in  each  direction,  but 
with  one  sum  differing  from  the  other. 

Peculiarities  found  in  the  dialects  of  one  village  will  reappear  in 
neighboring  localities,  but  there  is  nothing  to  show  exactly  how  far 
each  peculiarity  will  reach.  For  instance,  in  Douvaine,  a  locality  in 
the  department  of  Upper-Savoy,  the  name  of  Geneva  is  pro- 
nounced '^enva.  This  pronunciation  is  heard  far  to  the  east  and  to 
the  south,  but  on  the  other  side  of  Lake  Geneva  speakers  say 
dzenva.  Still,  it  is  not  a  question  of  two  clearly  distinct  dialects,  for 
the  boundaries  of  some  other  phenomenon  would  be  different.  In 
Douvaine,  speakers  say  daue  for  deux  'two,'  but  this  pronunciation 
has  a  much  more  restricted  zone  than  '6enva.  At  the  foot  of  the 
Sal^ve,  a  few  kilometers  away,  speakers  say  due. 

3.  Dialects  Have  No  Natural  Boundaries 

The  current  practice,  which  differs  from  ours,  is  to  picture  dia- 
lects as  perfectly  defined  linguistic  types,  bounded  in  all  directions 
and  covering  distinct  zones  placed  side  by  side  on  a  map  (a,  b,  c,  d, 
etc.).  But  natural  dialectal  transformations  produce  entirely  differ- 
ent results.  As  soon  as  we  studied  each  phenomenon  separately  and 
determined  its  spread,  our  old  notion  had  to  give  way  to  the  new 
one:  there  are  only  natural  dialectal  features,  not  natural  dialects; 
in  other  words,  there  are  as  many  dialects  as  there  are  localities. 

*  See  page  156.  [Tr.] 


r-M'  ^  V--H 

The  notion  of  natural  dialects  is  therefore  incompatible  with  the 
notion  of  fixed  well-defined  zones.  This  leaves  us  with  two  choices: 
(1)  we  may  define  a  dialect  by  the  totality  of  its  characteristics — 
which  involves  choosing  one  point  on  the  map  and  encompassing 
only  the  regional  speech-forms  of  a  single  locality  since  the  same 
peculiarities  will  not  extend  beyond  this  point ;  or  (2)  we  may  define 
a  dialect  by  one  of  its  characteristics,  and  simply  map  the  spread 
of  this  characteristic — which  obviously  is  an  artificial  procedure 
since  the  boundaries  that  we  mark  off  correspond  to  no  dialectal 

Research  in  dialectal  characteristics  was  the  point  of  departure 
for  works  on  linguistic  cartography.  The  model  linguistic  atlas  is 
GiUi^ron's  Atlas  linguistique  de  la  France.  Wenker's  map  of  Ger- 
many should  also  be  mentioned.^  The  form  of  the  atlas  is  predeter- 
mined, for  we  have  to  study  a  country  region  by  region,  and  a  map 
includes  only  a  small  number  of  the  dialectal  characteristics  of 
each  region.  One  must  sift  the  facts  for  each  region  many  times  to 
bring  to  light  the  phonetic,  lexicological,  morphological,  etc.  peculi- 
arities that  are  superimposed  on  each  other.  Such  an  undertaking 
requires  a  staff  of  experts,  well-planned  questionnaires,  the  co- 
operation of  local  correspondents,  etc.  One  noteworthy  project  is 
the  investigation  of  the  patois  of  French-speaking  Switzerland. 
Linguistic  atlases  are  useful  in  that  they  furnish  material  for  works 
on  dialectology.  Many  recent  monographs  are  based  on  Gillieron's 

The  boundaries  of  dialectal  characteristics  have  been  called 
isogloss  lines  or  isoglosses.  This  name,  coined  on  the  model  of 
isotherme,  is  obscure  and  inappropriate,  for  it  means  'having  the 
same  language.'  Since  glosseme  means  'idiomatic  character,'  the 

*  Cf.  also  Weigand,  Linguistischer  Atlas  des  dakorumdnischen  Gebiets  (1909) 
and  Millardet,  Petit  atlas  linguistique  d'une  region  des  Landes  (1910).   [S.] 


expression  isoglossematic  lines,  if  practical,  would  be  more  ap- 
propriate. But  I  prefer  to  use  innovating  waves,  a  descriptive  ex- 
pression that  goes  back  to  J.  Schmidt.  Chapter  III  will  show  the 
reasons  for  my  preference. 

A  glance  at  a  linguistic  atlas  will  sometimes  reveal  two  or  three 
waves  that  almost  coincide  or  even  overlap  in  one  zone : 

,'• — ir-'C-,.      

The  two  points  A  and  B,  which  are  separated  by  such  a  zone,  ob- 
viously have  some  divergencies  and  constitute  two  rather  clearly 
differentiated  forms  of  speech.  These  concordances,  instead  of 
being  partial,  may  characterize  the  whole  perimeter  of  two  or  more 


♦>.*x  t  'J 

A  dialect  is  defined,  roughly  speaking,  by  a  sufficient  accumulation 
of  such  concordances.  Their  foundations  are  social,  political,  re- 
ligious, etc.,  matters  which  do  not  concern  us  at  the  moment  but 
which  veil,  without  ever  erasing  completely,  the  basic  and  natural 
fact  of  differentiation  from  zone  to  zone. 

4.  Languages  Have  No  Natural  Boundaries 

Precisely  how  a  language  differs  from  a  dialect  is  hard  to  specify. 
Often  a  dialect  is  called  a  language  because  it  has  produced  a 
Uterature.  This  is  true  of  Portuguese  and  Dutch.  Intelligibility  also 
plays  a  part;  everyone  would  agree  that  people  who  do  not  under- 


stand  each  other  speak  different  languages.  Still,  languages  that 
have  evolved  over  continuous  territory  and  among  stable  popu- 
lations exhibit,  on  a  broader  scale,  the  same  facts  as  dialects. 
Innovating  waves  appear  here  too,  but  with  this  difference:  they 
embrace  a  zone  common  to  several  languages. 

It  is  impossible,  even  in  our  hypothetical  examples,  to  set  up 
boundaries  between  dialects.  The  same  applies  to  related  lan- 
guages. The  size  of  the  territory  makes  no  difference.  We  would  be 
unable  to  say  w^here  High  German  begins  and  Low  German  ends, 
and  would  find  it  just  as  impossible  to  draw  the  dividing  line  be- 
tween German  and  Dutch,  or  between  French  and  Italian.  There 
are  extreme  points  where  we  may  assert,  "Here  French  predomi- 
nates, here  Italian,"  but  in  the  intermediate  regions  the  distinction 
would  disappear.  We  might  imagine  a  compact,  more  restricted 
zone  of  transition  between  two  languages — e.g.  Frovengal  between 
French  and  Italian — but  such  a  zone  simply  does  not  exist.  How 
can  we  possibly  depict  an  exact  linguistic  boundary  on  territory 
that  is  covered  from  one  end  to  the  other  by  gradually  differ- 
entiated dialects?  The  dividing  lines  between  languages,  hke  those 
between  dialects,  are  hidden  in  transitions.  Just  as  dialects  are  only 
arbitrary  subdivisions  of  the  total  surface  of  language,  so  the 
boundary  that  is  supposed  to  separate  two  languages  is  only  a 
conventional  one. 

Still,  abrupt  transitions  from  one  language  to  another  are  com- 
mon, due  to  circumstances  that  have  destroyed  imperceptible  tran- 
sitions. The  most  disrupting  force  is  the  shifting  of  populations. 
Nations  have  always  shuttled  back  and  forth.  Their  migrations, 
multiplied  throughout  the  centuries,  have  wrought  confusion 
everywhere,  and  at  many  points  all  trace  of  linguistic  transition 
has  been  wiped  out.  The  Indo-European  family  is  typical.  At  first 
its  languages  must  have  been  closely  related,  with  an  unbroken 
chain  of  linguistic  zones.  We  can  reconstruct  the  broad  outlines  of 
the  major  zones.  Slavic  shares  overlapping  characteristics  with 
both  Iranian  and  Germanic,  and  this  conforms  with  the  geographi- 
cal distribution  of  the  three  languages;  similarly,  Germanic  is  an 
intermediate  ring  that  links  Slavic  and  Celtic,  which  in  turn  is 
closely  related  to  Italic ;  the  latter  is  mid-way  between  Celtic  and 
Greek.  Thus  a  Unguist,  without  knowing  its  geographical  location, 


could  readily  assign  each  idiom  to  its  proper  place.  And  yet,  as  soon 
as  we  consider  a  boundary  between  two  groups  of  idioms  (e.g. 
the  Germanic-Slavic  boundary),  there  is  an  abrupt  break,  with  no 
transition.  The  two  groups  colUde  instead  of  overlapping.  That  is 
because  the  intermediate  dialects  have  disappeared.  Neither  the 
Slavs  nor  the  Germans  were  stationary;  they  emigrated,  conquered 
territory,  each  at  the  expense  of  the  other ;  the  neighboring  Slavic 
and  Germanic  populations  of  today  are  not  the  same  as  those  that 
were  once  in  contact.  If  the  Italians  who  live  in  Calabria  settled  on 
the  French  border,  the  move  would  naturally  destroy  the  im- 
perceptible transition  between  Italian  and  French.  A  number  of 
similar  facts  accounts  for  the  distribution  of  Proto-Indo-European. 
Still  other  forces  help  to  wipe  out  transitions.  Take  the  spreading 
of  standard  languages  at  the  expense  of  patois  (see  pp.  195  ff.).  To- 
day literary  French  (formerly  the  language  of  the  He  de  France) 
extends  to  the  border,  where  it  conflicts  with  official  Italian  (a 
generalized  form  of  the  Tuscan  dialect),  and  it  is  only  through 
chance  that  traditional  patois  still  exist  in  the  western  Alps,  for 
along  many  other  linguistic  boundaries  all  trace  of  intermediate 
speech-forms  has  been  wiped  out. 

Chapter  IV 

1.  Intercourse^  and  Provincialism 

The  laws  that  govern  the  spread  of  linguistic  phenomena  are  the 
same  as  those  that  govern  any  custom  whatsoever,  e.g.  fashion.  In 
every  human  collectivity  two  forces  are  always  working  simul- 
taneously and  in  opposing  directions:  individualism  or  provincial- 
ism [esprit  de  clocher]  on  the  one  hand  and  intercourse — communi- 
cations among  men — on  the  other. 

Provincialism  keeps  a  restricted  linguistic  community  faithful 
to  its  own  traditions.  The  patterns  that  the  individual  acquires 

*  In  his  lectures  Saussure  used  the  English  word  intercourse.  [Tr.] 


during  childhood  are  strong  and  persistent.  If  they  alone  were  at 
work,  these  patterns  would  create  an  infinite  number  of  peculi- 
arities in  speech. 

But  intercourse,  the  opposing  force,  limits  their  effect.  Whereas 
provincialism  makes  men  sedentary,  intercourse  obliges  them  to 
move  about.  Intercourse  brings  passers-by  from  other  localities 
into  a  village,  displaces  a  part  of  the  population  whenever  there  is 
a  festival  or  fair,  unites  men  from  different  provinces  in  the  army, 
etc.  In  a  word,  it  is  a  unif3ang  force  that  counteracts  the  splintering 
action  of  provincialism. 

Intercourse  spreads  language  and  gives  it  unity.  It  acts  in  two 
ways:  negatively,  it  prevents  dialectal  splintering  by  wiping  out 
an  innovation  whenever  and  wherever  it  springs  up ;  positively,  it 
promotes  unity  by  adopting  and  spreading  an  innovation.  The 
second  form  that  intercourse  may  take  justifies  the  use  of  the  word 
wave  to  designate  the  geographical  boundaries  of  a  dialectal  fact 
(see  p.  203),  for  an  isoglossematic  line  is  like  the  outermost  edge 
of  an  undulating  flood. 

Surprisingly  enough,  we  sometimes  find  that  two  widely  sepa- 
rated dialects  within  the  same  language  have  a  common  linguistic 
trait.  That  is  because  the  change  which  sprang  up  at  one  place  on 
the  territory  met  no  obstacle  in  spreading  and  gradually  extended 
far  beyond  its  starting  point.  Nothing  impedes  the  action  of  inter- 
course in  a  linguistic  mass  within  which  there  are  only  imper- 
ceptible transitions. 

The  generahzing  of  a  particular  fact — regardless  of  the  size  of 
its  zone — requires  time,  and  occasionally  the  time  is  measurable. 
Thus  the  change  of  ])  to  d,  which  intercourse  carried  throughout 
continental  Germany,  first  spread  over  the  south,  between  800  and 
850  A.D.,  except  for  Franconia  where  ])  persisted  as  soft  6  and  did 
not  give  way  to  d  until  a  later  date.  The  change  of  t  to  German  z 
(pronounced  ts)  took  place  within  more  restricted  boundaries  and 
began  during  a  period  that  preceded  the  first  written  documents; 
it  must  have  started  in  the  Alps  around  600  a.d.  and  spread  both 
north  and  south  as  far  as  Lombardy.  The  t  still  appears  in  an 
eighth-century  Thuringian  charter.  During  a  later  period  Germanic 
I  and  u  were  diphthongized  (cf .  mein  for  mln,  braun  for  hriln) ;  it 
took  300  years  for  this  phenomenon,  which  began  in  Bohemia 


around  1400  a.d.,  to  reach  the  Rhine  and  cover  its  present  zone. 

The  foregoing  Hnguistic  facts  spread  through  interdialectal 
influence,  and  the  same  is  probably  true  of  all  waves:  they  start 
from  one  point  and  radiate.  This  brings  us  to  a  second  important 

German  consonantal  mutation  is  again  illustrative.  When  the 
phoneme  t  became  ts  at  one  point  in  Germanic  territory,  the  new 
sound  tended  to  radiate  from  its  source,  and  ts  became  the  rival  of 
the  original  t  or  of  other  sounds  that  might  have  evolved  from  it 
at  other  points.  At  its  source  such  an  innovation  is  purely  phonetic, 
but  elsewhere  it  becomes  estabhshed  only  geographically  and 
through  interdialectal  influence.  Hence  the  diagram : 



is  valid  in  all  its  simphcity  for  the  source  and  no  more.  If  we  try 
to  apply  it  to  propagation,  the  resulting  picture  is  distorted. 

The  phonetician  must  therefore  distinguish  carefully  between 
sources  and  affected  zones.  At  its  source  a  phoneme  evolves  solely 
on  the  axis  of  time.  But  mere  phonetic  facts  will  not  explain 
affected  zones,  for  they  result  from  the  interaction  of  both  time 
and  space.  Take  ts,  which  came  from  an  outside  source  and  replaced 
t.  This  is  an  example,  not  of  modification  of  a  traditional  prototype, 
but  of  imitation  of  a  neighboring  dialect,  irrespective  of  the  proto- 
type. Herza  'heart'  came  from  the  Alps  and  replaced  the  more 
archaic  form  herta  in  Thuringia.  Here  we  should  not  speak  of 
phonetic  change  but  of  the  borrowing  of  a  phoneme. 

2.  The  Two  Forces  Reduced  to  One 

If  we  focus  on  a  single  geographical  point — by  "point"  I  mean 
a  very  small  area  comparable  to  a  point  (see  p.  202),  e.g.  a  village — 
it  is  easy  to  single  out  what  is  attributable  to  each  of  the  two 
forces,  provincialism  and  intercourse.  Any  particular  fact  depends 
on  only  one  force,  never  on  both;  every  feature  shared  with  another 
dialect  is  due  to  intercourse ;  every  feature  that  belongs  exclusively 
to  the  dialect  of  the  point  under  consideration  is  due  to  pro- 


But  as  soon  as  we  turn  to  a  larger  area — e.g.  a  canton — a  new 
difficulty  arises.  No  longer  is  it  possible  to  say  which  force  is  re- 
sponsible for  a  given  phenomenon.  Both  forces,  though  in  oppo- 
sition, are  involved  in  each  trait  of  the  idiom.  What  is  distinctive 
of  canton  A  is  common  to  all  its  parts.  There,  the  individualistic 
force  prohibits  canton  A  from  imitating  something  from  canton  B 
and  the  latter  in  turn  from  imitating  A.  But  the  unifying  force, 
intercourse,  is  also  involved,  for  it  shows  up  in  the  different  parts 
of  A  (A\  A^,  A^,  etc.).  On  larger  areas  the  two  forces  therefore  work 
simultaneously  but  in  different  proportions.  The  more  intercourse 
favors  an  innovation,  the  farther  its  zone  will  reach;  as  for  pro- 
vincialism, it  tends  to  protect  a  linguistic  fact  throughout  its  zone 
by  defending  it  against  outside  competitors.  We  cannot  foresee  the 
final  results  of  the  action  of  the  two  forces.  In  Germanic  territory, 
which  reached  from  the  Alps  to  the  North  Sea,  the  change  from  ]> 
to  d  was  general  while  the  change  from  t  to  is  affected  only  the 
south  (see  p.  206) ;  provincialism  created  an  opposition  between 
the  south  and  the  north,  but  intercourse  was  responsible  for  lin- 
guistic solidarity  within  each  region.  Thus  there  is  basically  no 
difference  between  this  second  phenomenon  and  the  first.  The  same 
forces  are  present;  only  the  intensity  of  their  action  varies. 

Practically,  this  means  that  in  studying  linguistic  evolutions  we 
can  disregard  the  individualistic  force.  That  is,  we  can  consider  it 
as  the  negative  side  of  the  unifying  force.  The  latter  may  be  strong 
enough  to  unify  the  whole  area.  If  not,  the  phenomenon  will  come 
to  a  standstill  after  covering  only  a  part  of  the  territory.  Internally, 
however,  the  part  that  was  covered  will  form  a  coherent  whole. 
That  is  why  we  can  reduce  everything  to  the  single  unifying  force 
without  bringing  in  provincialism,  which  is  nothing  more  than  the 
force  of  intercourse  peculiar  to  each  region. 

3.  Linguistic  Differentiation  on  Separate  Territories 

Three  things  must  be  realized  before  one  can  study  profitably  a 
language  that  develops  concurrently  on  two  separate  territories: 
(1)  in  a  unilingual  mass  cohesiveness  is  not  the  same  for  all  phe- 
nomena; (2)  not  all  innovations  spread;  and  (3)  geographical  con- 
tinuity does  not  prevent  perpetual  differentiations. 

Such   concurrent   development  is   common.   When   Germanic 


crossed  over  from  the  continent  to  the  British  Isles,  for  example, 
there  began  a  twofold  evolution.  On  the  one  hand  were  the  German 
dialects  and  on  the  other  Anglo-Saxon,  from  which  English 
evolved.  Another  example  is  French  after  it  was  transplanted  to 
Canada.  Discontinuity  is  not  always  the  effect  of  colonization  or 
conquest;  it  may  also  result  from  isolation.  Rumanian  lost  contact 
with  the  Latin  mass  through  the  interposition  of  Slavic  popu- 
lations. The  cause  is  unimportant ;  what  matters  is  whether  sepa- 
ration plays  a  role  in  the  history  of  languages  and  whether  its 
effects  differ  from  those  that  appear  where  there  is  continuity. 

Earlier,  in  order  to  point  up  the  preponderant  effect  of  time,  we 
imagined  an  idiom  as  it  might  develop  concurrently  on  two  rather 
limited  points — two  small  islands,  in  our  example — where  we  might 
disregard  a  gradual  spread.  Now,  however,  with  two  territories 
that  cover  a  broader  area,  we  find  once  more  that  a  gradual  spread 
brings  about  dialectal  differences.  That  the  two  territories  are  dis- 
continuous does  not  simplify  the  problem  in  the  least.  We  must 
guard  against  attributing  to  separation  something  that  can  be 
explained  without  it. 

This  is  the  mistake  that  the  earliest  Indo-European  scholars 
made  (see  p.  2).  Confronted  with  a  great  family  of  languages  that 
had  diverged  enormously,  they  failed  to  realize  that  the  differences 
could  have  resulted  from  something  besides  geographical  splinter- 
ing. It  was  easy  for  them — and  for  anyone — to  imagine  different 
languages  in  separate  localities;  in  a  superficial  view  no  more  was 
needed  to  explain  differentiation.  But  they  went  further.  They 
associated  nationality  with  language,  using  the  first  to  explain  the 
second.  Thus  they  pictured  the  Slavs,  Germans,  Celts,  etc.  as  so 
many  swarms  of  bees  from  the  same  hive  and  imagined  that  these 
tribes,  torn  away  from  the  original  stock  by  migration,  had  carried 
Proto-Indo-European  over  as  many  different  territories. 

Only  much  later  was  this  mistake  corrected.  Not  until  1877  did 
Johannes  Schmidt  open  the  eyes  of  linguists  by  proposing  the 
theory  of  continuity  or  waves  (Wellentheorie)  in  his  book  Die  Ver- 
wandtschaftsverhdltnisse  der  Indogermanen.  Then  they  saw  that 
local  splintering  suffices  to  explain  the  reciprocal  relations  of  the 
Indo-European  languages,  and  that  it  is  not  necessary  to  assume 
that  the  different  nations  moved  to  new  places  (see  p.  204).  Dia- 


lectal  differentiations  could  and  must  have  arisen  before  these 
nations  spread  out  in  various  directions.  The  wave  theory  there- 
fore not  only  gives  a  truer  picture  of  Proto-Indo-European ;  it  also 
reveals  the  causes  of  differentiation  and  the  conditions  that  de- 
termine the  kinship  of  languages. 

The  wave  theory  opposes  the  migratory  theory  but  does  not 
necessarily  exclude  it.  In  the  history  of  the  Indo-European  lan- 
guages there  are  many  examples  of  nations  that  lost  contact  with 
the  main  family  through  migration,  and  this  must  have  produced 
special  effects.  But  these  effects  mingle  with  those  of  differentiation 
where  contact  is  maintained,  and  the  difficulty  of  identifying  them 
brings  us  back  to  the  problem  of  the  evolution  of  an  idiom  in  sepa- 
rate territories. 

Take  Old  EngUsh.  It  broke  away  from  the  Germanic  trunk  as  a 
result  of  migration.  In  all  probabihty  it  would  not  have  its  present 
form  if  the  Saxons  had  stayed  on  the  continent  during  the  fifth 
century.  But  what  were  the  specific  effects  of  separation?  It  would 
seem  that  we  should  first  ask  whether  such  and  such  a  change 
might  not  have  sprung  up  just  as  well  where  geographical  contact 
was  maintained.  If  the  English  had  occupied  Jutland  instead  of 
the  British  Isles,  it  is  possible  that  some  of  the  facts  attributed  to 
absolute  separation  would  have  occurred  here  in  a  contiguous 
territory.  There  is  nothing  to  prove  that  discontinuity  is  what 
enabled  English  to  preserve  older  ]?  while  the  sound  became  d 
throughout  the  continent  (e.g.  English  thing  and  German  Ding). 
Nor  was  geographical  continuity  necessarily  responsible  for  the 
generalizing  of  the  change  in  continental  Germanic ;  it  might  very 
well  have  been  checked  in  spite  of  continuity.  The  mistake  is  the 
usual  one  of  contrasting  isolated  and  continuous  dialects.  Nothing 
actually  proves  that  interdialectal  influence  would  have  caused  d 
to  spread  throughout  our  imaginary  English  colony  in  Jutland.  We 
have  seen  that  in  the  linguistic  territory  of  French,  for  example, 
k  {-\-  a)  persisted  in  the  angle  formed  by  Picardy  and  Normandy 
but  became  hushing  h  (ch)  everywhere  else.  Isolation  is  therefore 
an  unsatisfactory  and  superficial  explanation.  Differentiation  can 
always  be  explained  without  it.  What  isolation  can  do,  geo- 
graphical continuity  does  equally  well.  If  there  is  a  difference 
between  the  two  classes  of  phenomena,  we  cannot  grasp  it. 


But  the  picture  changes  when  we  consider  two  related  idioms 
not  from  the  negative  viewpoint  of  their  differences  but  from  the 
positive  viewpoint  of  their  sohdarity.  Then  we  see  that  separation 
immediately  opens  the  door  to  potential  severance  of  every  relation 
whereas  geographical  continuity  supports  solidarity  even  among 
strikingly  different  regional  speech-forms,  provided  they  are 
connected  by  intermediate  dialects. 

In  order  to  determine  degrees  of  kinship  among  languages,  we 
must  therefore  make  a  rigid  distinction  between  continuity  and 
isolation.  Two  isolated  languages  will  retain  from  their  common 
heritage  a  number  of  traits  that  attest  their  kinship,  but  since  each 
language  will  evolve  independently,  new  characteristics  that 
appear  in  one  will  not  be  found  in  the  other  (with  the  exception  of 
certain  characteristics  that  originate  after  separation  and  are 
identical  in  the  two  languages  through  sheer  coincidence).  What 
is  ruled  out  in  each  instance  is  the  spreading  of  these  charac- 
teristics through  interdialectal  influence.  A  language  that  has 
evolved  out  of  touch  with  related  languages  generally  has  a  set  of 
traits  that  distinguish  it  from  them.  When  this  language  splinters 
in  turn,  its  dialects  evidence  a  closer  kinship  through  the  common 
traits  that  bind  them  together  and  set  them  apart  from  dialects  of 
the  other  territory.  They  actually  form  a  distinct  branch,  detached 
from  the  trunk. 

Vastly  different  are  the  relations  of  languages  on  continuous 
territory.  Their  common  traits  are  not  necessarily  older  than  the 
traits  that  differentiate  them.  Indeed,  an  innovation  that  starts  at 
a  given  point  may  spread  at  any  moment  and  even  embrace  the 
whole  territory.  Besides,  innovating  zones  vary  in  extent,  so  that 
two  neighboring  idioms  may  have  a  common  peculiarity  without 
forming  a  separate  group,  and  each  may  be  related  to  contiguous 
idioms  through  other  traits,  as  is  shown  by  the  Indo-European 


Concerning  Retrospective 

Chapter  I 


Synchronic  linguistics  has  only  the  perspective  of  speakers  and, 
consequently,  only  one  method;  diachronic  linguistics,  however, 
requires  both  a  prospective  and  a  retrospective  viewpoint  (see 
p.  90). 

The  prospective  method,  which  corresponds  to  the  actual  course 
of  events,  is  the  one  we  must  use  in  developing  any  point  concern- 
ing the  history  of  a  language  or  of  languages.  It  consists  simply  of 
examining  the  available  documents.  But  all  too  many  problems  of 
diachronic  linguistics  cannot  be  met  by  the  prospective  method. 

In  fact,  in  order  to  give  a  detailed  history  of  a  language  by  fol- 
lowing its  course  in  time,  one  would  need  an  infinite  number  of 
photographs,  taken  at  different  times.  Now  this  requirement  has 
never  been  met.  Romance  scholars,  for  instance,  even  though  they 
have  the  advantage  of  knowing  Latin,  the  point  of  departure  for 
their  research,  and  of  possessing  an  imposing  array  of  documents 
covering  several  successive  centuries,  are  constantly  aware  of  wide 
gaps  in  their  documentation.  They  must  then  discard  the  pro- 
spective method — direct  evidence — and  work  in  the  opposite 
direction,  using  the  retrospective  method  to  retrace  time.  This 
means  choosing  a  particular  period  and  trying  to  determine,  not 
how  a  form  developed,  but  the  oldest  form  that  could  have  given 
it  birth. 

The  prospective  method  amounts  to  simple  narration  and  is 
based  entirely  on  textual  criticism,  but  the  retrospective  viewpomt 
requires  a  reconstructive  method  supported  by  comparison.  It  is 



impossible  to  establish  the  original  form  of  a  single,  isolated  sign, 
but  the  comparing  of  two  different  signs  that  have  the  same  origin 
(e.g.  Latin  pater,  Sanskrit  pilar-  or  the  radical  of  Latin  ger-o  and 
that  of  ges-tus)  immediately  brings  to  hght  the  diachronic  unity 
which  relates  both  signs  to  a  prototype  that  can  be  reconstructed 
inductively.  The  more  numerous  the  comparisons,  the  more  accu- 
rate inductions  will  be,  and  the  results — if  sufficient  data  are  at 
hand — will  be  true  reconstructions. 

The  same  applies  to  languages  in  their  totality.  We  can  infer 
nothing  about  Basque ;  because  it  is  isolated,  there  is  nothing  with 
which  we  can  compare  it.  But  by  comparing  a  group  of  related  lan- 
guages like  Greek,  Latin,  Old  Slavic,  etc.,  scholars  were  able  to 
single  out  the  common  original  elements  and  to  reconstruct  the 
essentials  of  Proto-Indo-European  as  it  existed  before  differenti- 
ation in  space  occurred.  What  was  done  for  the  whole  family  on  a 
large  scale  was  repeated  on  a  smaller  scale — and  always  by  the 
same  procedure — for  each  of  its  parts  wherever  this  was  necessary 
and  possible.  We  know  numerous  Germanic  idioms  directly, 
through  documents,  but  we  know  Proto-Germanic — the  source  of 
these  different  idioms — only  indirectly,  through  the  reconstructive 
method.  Using  the  same  method  with  varying  success,  linguists 
have  also  sought  the  original  unity  of  other  families  (see  p.  192). 

The  retrospective  method,  then,  takes  us  far  beyond  the  oldest 
documents  in  tracing  the  history  of  a  language.  Thus  it  was  pos- 
sible to  draw  the  prospective  outline  of  Latin,  whose  history  hardly 
begins  before  the  third  or  fourth  century  B.C.,  only  after  the  re- 
construction of  Proto-Indo-European  had  given  an  inkling  of  what 
must  have  happened  between  the  period  of  original  unity  and  the 
first  known  Latin  documents. 

With  respect  to  reconstruction,  evolutionary  linguistics  is  like 
geology,  another  historical  science.  Geology  sometimes  has  to 
describe  stable  states  (e.g.  the  present  state  of  Lake  Geneva  Basin) 
without  considering  what  might  have  preceded  in  time,  but  its 
main  concern  is  the  chain  of  events  and  transformations  that  make 
up  diachronics.  A  prospective  geology  is  conceivable,  but  in  reality 
the  viewpoint  is  usually  only  retrospective.  Before  recounting 
what  has  occurred  at  a  given  point  on  the  earth,  the  geologist  must 


reconstruct  the  chain  of  events  and  try  to  determine  what  is 
responsible  for  the  present  state  of  that  part  of  the  globe. 

Not  only  in  method  do  the  two  perspectives  contrast  sharply; 
in  teaching,  even,  to  use  them  simultaneously  in  the  same  expo- 
sition is  a  disadvantage.  The  study  of  phonetic  changes,  for  in- 
stance, offers  two  very  different  pictures,  depending  on  the 
perspective.  Using  the  prospective  viewpoint,  we  might  ask  what 
Classical  Latin  e  became  in  French.  We  would  see  that  a  single 
sound,  by  evolving  in  time,  varied  and  gave  rise  to  several  pho- 
nemes: cf.  pedem  -^  pye  (pied  'foot'),  ventum  — ^  vd  (vent  'wind'), 
lectum  -^  li  (lit  'bed'),  necdre  -^  nwaye  (noyer  'drown'),  etc. 
Against  that,  if  we  used  the  retrospective  viewpoint  to  find  what 
French  open  e  stands  for  in  Latin,  we  would  see  that  this  single 
sound  is  the  terminal  point  of  several  originally  distinct  phonemes : 
cf.  ter  (terre  'earth')  =  terram,  verz  (verge  'rod')  =  virgam,  fe  (fait 
'fact')  =  factum,  etc.  We  could  present  the  evolution  of  formative 
elements  in  two  ways,  and  the  two  pictures  would  be  just  as  differ- 
ent; everything  that  was  said  about  analogical  formations  (see 
pp.  169  ff.)  is  a  priori  proof.  Thus  the  (retrospective)  search  for  the 
origin  of  the  suffix  of  French  participles  in  -e  takes  us  back  to  Latin 
-dtum;  the  Latin  suffix  is  related  etymologically  to  denominative 
Latin  verbs  in  -are,  which  go  back  mainly  to  feminine  substantives 
in  -a  (cf.  plantdre:  planta,  Greek  tlmad:  tlma,  etc.);  furthermore, 
-dtum  would  not  exist  if  the  Proto-Indo-European  suffix  -to-  had 
not  been  living  and  productive  in  its  own  right  (cf.  Greek  klu-to-s, 
Latin  in-clu-tu-s,  Sanskrit  gru-ta-s,  etc.) ;  finally,  -dtum  includes  the 
formative  element  -m  of  the  accusative  singular  (see  p.  154). 
Conversely,  a  (prospective)  search  for  the  French  formations  that 
have  the  original  suffix  -to-  will  reveal  that  there  are  not  only 
the  different  sufl&xes — whether  productive  or  not — of  the  past 
participle  (aime  'loved'  =  amdtum,  Jini  'ended'  =  finltum,  clos 
'closed'  =  clausum  for  *claudtum,  etc.),  but  also  many  others  like 
-u  =  -utum  (cf.  cornu  'horned'  =  cornutum),  -tif  (learned  suffix) 
=  Latin  -tivum  (cf .  fugitif  =  fugitivum,  sensitif,  negatif,  etc.)  and 
a  number  of  words  no  longer  analyzable,  like  point  'dot'  =  Latin 
punctum,  de  'die'  =  datum,  chetif  'wretched'  =  captlvum,  etc. 


Chapter  II 


In  the  earliest  stages  of  Indo-European  linguistics  scholars  under- 
stood neither  the  real  purpose  of  comparison  nor  the  importance 
of  the  reconstructive  method  (see  p.  3).  That  explains  one  of 
their  grossest  mistakes:  the  exaggerated  and  almost  exclusive  role 
that  they  gave  to  Sanskrit.  Because  it  was  the  oldest  document  of 
Proto-Indo-European,  they  promoted  Sanskrit  to  the  rank  of 
prototype.  To  imagine  that  Proto-Indo-European  engendered 
Sanskrit,  Greek,  Slavic,  Celtic,  Italic,  etc.  is  one  thing;  to  sub- 
stitute one  of  these  languages  for  Proto-Indo-European  is  some- 
thing else  entirely.  The  glaring  mistake  of  the  earliest  scholars  had 
varied  and  far-reaching  consequences.  Doubtless  their  hypothesis 
was  not  stated  so  categorically  as  I  have  implied,  but  it  was  tacitly 
accepted  in  practice.  Bopp  wrote  that  he  "did  not  think  that 
Sanskrit  could  be  the  common  source,"  as  if  there  were  a  possibility 
of  formulating,  even  while  expressing  doubt,  such  a  supposition. 
This  prompts  one  to  ask  what  is  meant  by  the  statement  that 
one  language  is  older  than  another.  Three  interpretations  are 
theoretically  possible : 

(1)  "Older"  may  refer  to  the  beginning,  the  starting  point  of  a 
language.  But  only  a  little  reasoning  will  show  that  there  is  no 
language  to  which  we  can  assign  an  age,  for  each  language  is  the 
continuation  of  what  was  spoken  before  it.  What  is  true  of  hu- 
manity is  not  true  of  speech ;  the  absolute  continuity  of  its  develop- 
ment prevents  us  from  distinguishing  generations  in  it.  Gaston 
Paris  was  justified  in  criticizing  the  conception  of  daughter  lan- 
guages and  mother  languages  since  this  assumes  interruptions. 
"Older,"  in  this  sense,  is  meaningless. 

(2)  "Older"  may  also  indicate  that  one  particular  state  of  a 
language  we  are  studying  is  earlier  than  another  state  of  the  same 
language.  Thus  the  Persian  of  the  Achaemenian  inscriptions  is 


older  than  the  Persian  of  Firdausi.  In  a  specific  case  Uke  this,  where 
one  idiom  has  definitely  developed  from  the  other  and  where  both 
are  equally  well  known,  we  should  of  course  reckon  only  with  the 
earlier  idiom.  But  unless  both  conditions  are  met,  priority  in  time 
has  no  importance.  Thus  Lithuanian,  which  is  attested  only  since 
1540,  is  no  less  valuable  than  Old  Slavic,  which  was  recorded  in  the 
tenth  century,  or  than  the  Sanskrit  of  the  Rig  Veda  for  that 

(3)  Finally,  "older"  may  designate  a  more  archaic  language- 
state,  i.e.  one  with  forms  that  are  very  close  to  the  forms  of  the 
original  model,  quite  apart  from  any  question  of  dates.  In  this 
sense  sixteenth-century  Lithuanian  is  older  than  the  Latin  of  the 
third  century  B.C. 

Only  in  the  second  or  third  sense  is  Sanskrit  older  than  other 
languages.  It  fits  both  definitions.  On  one  hand,  it  is  generally 
agreed  that  the  Vedic  hymns  antedate  the  oldest  Greek  texts;  on 
the  other  hand — and  this  is  especially  important — Sanskrit  has  a 
considerable  number  of  archaic  features  in  comparison  with  those 
preserved  by  other  languages  (see  pp.  2  ff.). 

But  the  earliest  linguists,  because  of  their  confused  notion  of  age, 
put  Sanskrit  ahead  of  the  whole  family.  The  result  was  that  later 
linguists,  though  cured  of  the  notion  that  Sanskrit  is  the  mother 
language,  continued  to  attribute  too  much  importance  to  the 
evidence  that  it  furnishes  as  a  collateral  language. 

In  Les  Origines  indo-europeennes  (see  p.  224)  A.  Pictet,  while 
explicitly  recognizing  the  existence  of  a  primitive  nation  with  its 
own  language,  still  insists  that  we  must  first  consult  Sanskrit,  and 
that  the  evidence  which  this  language  furnishes  is  worth  more  than 
that  of  several  other  Indo-European  languages  combined.  The 
same  delusion  has  for  many  years  obscured  issues  of  primary 
importance,  such  as  that  of  the  Proto-Indo-European  vocaUsm. 

The  mistake  has  been  repeated  on  a  smaller  scale  and  in  detail. 
Those  who  studied  specific  branches  of  Indo-European  thought 
that  the  earliest  known  idiom  was  a  complete  and  satisfactory 
representative  of  the  whole  group  and  did  not  try  to  become  better 
acquainted  with  the  original  state.  For  example,  instead  of  speak- 
ing of  Germanic,  they  had  no  scruples  about  citing  Gothic  and 
stopping  there,  for  Gothic  antedates  the  other  Germanic  dialects 


by  several  centuries ;  it  usurped  the  role  of  prototype  and  became 
the  source  of  the  other  dialects.  As  regards  Slavic,  they  based  their 
research  exclusively  on  Slavonic  or  Old  Slavic,  which  is  attested 
from  the  tenth  century,  because  the  other  Slavic  dialects  are 
attested  from  a  later  date. 

Only  on  very  rare  occasions  do  two  specimens  of  language  that 
have  been  set  down  in  writing  at  successive  dates  represent  exactly 
the  same  idiom  at  two  moments  in  its  history.  More  often  we  find 
that  one  of  the  dialects  is  not  the  linguistic  successor  of  the  other. 
Exceptions  prove  the  rule.  The  most  famous  exception  is  the 
Romance  languages  with  respect  to  Latin :  in  tracing  French  back 
to  Latin,  one  certainly  follows  a  vertical  route;  the  territory  of  the 
Romance  languages  happens  to  match  the  territory  where  Latin 
was  spoken,  and  each  idiom  is  no  more  than  a  later  state  of  Latin. 
Persian  is  another  exception  to  the  rule;  the  Persian  of  the  in- 
scriptions of  Darius  is  the  same  dialect  as  the  Persian  of  the  Middle 
Ages.  But  the  opposite  occurs  much  more  frequently.  The  written 
documents  of  different  periods  generally  belong  to  different  dia- 
lects of  the  same  family.  Germanic,  for  instance,  appears  succes- 
sively in  the  Gothic  of  Ulfilas  (its  successor  is  unknown),  then  in 
Old  High  German  texts,  later  in  Anglo-Saxon  and  Old  Norse  texts, 
etc.  None  of  these  dialects  or  groups  of  dialects  is  the  continuation 
of  the  one  attested  previously.  The  following  diagram,  in  which 
letters  stand  for  dialects  and  dotted  lines  for  successive  periods, 
suggests  the  usual  pattern : 

A . . . .  Period  1 

B Period  2 

..C D Periods 

E..  Period  4 

This  pattern  is  a  valuable  asset  to  linguistics.  If  succession  were 
vertical,  the  first  known  dialect  (A)  would  contain  everything  that 
we  could  deduce  by  analyzing  successive  states.  But  by  searching 
for  the  point  of  convergence  of  all  the  dialects  (A,  B,  C,  D,  etc.)  in 
the  pattern,  we  may  find  a  form  older  than  A  (i.e.  a  prototype  X) 
and  thus  avoid  confusing  A  and  X. 


Chapter  III 

1.  Their  Nature  and  Aim 

The  sole  means  of  reconstructing  is  by  comparing,  and  the  only 
aim  of  comparison  is  a  reconstruction.  Our  procedure  is  sterile 
unless  we  view  the  relations  of  several  forms  from  the  perspective 
of  time  and  succeed  in  re-establishing  a  single  form.  I  have  re- 
peatedly emphasized  this  point  (see  pp.  3  ff.  and  p.  198  f.).  Thus 
we  explain  Latin  medius  against  Greek  mesos,  without  going  back 
to  Proto-Indo-European,  by  positing  an  older  form  *methyos  as  the 
source  of  both  medius  and  mesos.  Or  we  may  compare  two  forms  of 
the  same  language  rather  than  two  words  of  different  languages: 
Latin  gero  and  gestus  go  back  to  a  radical  *ges-  that  was  once 
common  to  both  forms. 

We  note  in  passing  that  comparisons  having  to  do  with  phonetic 
changes  must  always  rely  heavily  on  morphological  considerations. 
In  examining  Latin  patior  and  passus,  I  bring  in  f actus,  dictus,  etc. 
because  passus  is  a  formation  of  the  same  class.  By  basing  my  con- 
clusion on  the  morphological  relation  between  facio  and  factus, 
died  and  dictus,  etc.,  I  can  set  up,  for  an  earlier  period,  the  same 
relation  between  patior  and  *pat-tus.  Reciprocally,  I  must  use 
phonetics  to  throw  light  on  a  morphological  comparison.  I  can 
compare  Latin  meliorem  with  Greek  hedio  because  the  first  form 
goes  back  phonetically  to  *meliosem,  *meliosm,  and  the  second  to 
*hadioa,  *hddiosa,  *hddiosm. 

Linguistic  comparison  is  not  simply  a  mechanical  operation.  It 
implies  the  bringing  together  of  all  relevant  data.  But  it  must 
always  result  in  a  conjecture  which  we  can  express  by  some  formula 
and  which  aims  to  re-establish  something  that  has  preceded;  it 
always  results  in  a  reconstruction  of  forms. 

But  is  the  aim  of  viewing  the  past  to  reconstruct  the  whole, 
concrete  forms  of  the  previous  state?  Or  is  reconstruction  limited 


to  abstract,  partial  affirmations  about  word-parts  (e.g.  to  the  ob- 
servation that  Latin  /  in  fumus  stands  for  Proto-Itahc  ]),  or  that 
the  initial  element  of  Greek  alio  and  Latin  aliud  already  existed  as 
a  in  Proto-Indo-European)  ?  Reconstruction  may  well  confine  itself 
to  the  second  type  of  research;  its  analytical  method  has  no  aim 
other  than  these  partial  observations.  Still,  from  the  sum  of  isolated 
facts,  we  can  draw  general  conclusions.  A  series  of  facts  similar  to 
those  pertainmg  to  fumus  allows  us  to  state  with  certainty  that  \> 
had  a  place  in  the  phonological  system  of  Proto-Italic ;  similarly, 
we  can  state  that  the  pronominal  declension  of  Proto-Indo-Euro- 
pean has  a  neuter  singular  ending  -d,  different  from  the  -m  of 
adjectives.  We  deduce  this  general  morphological  fact  from  a  set 
of  isolated  observations  (cf.  Latin  istud,  aliud  against  honum; 
Greek  to  =  *tod,  alio  =  *allod  against  kalon;  English  that,  etc.). 
We  can  go  even  further.  It  is  possible,  after  we  have  reconstructed 
the  different  facts,  to  synthesize  those  relating  to  the  whole  form 
and  to  reconstruct  whole  words  (e.g.  Proto-Indo-European  *alyod), 
inflectional  paradigms,  etc.  Synthesis  consists  of  drawing  together 
completely  isolated  statements.  For  example,  when  we  compare 
the  different  parts  of  a  reconstructed  form  like  *alyod,  we  notice  a 
great  difference  between  the  -d,  which  raises  a  point  of  grammar, 
and  a-,  which  has  no  grammatical  significance.  A  reconstructed 
form  is  not  a  solidary  whole.  It  is  a  sum  that  we  can  always  analyze 
phonetically.  Each  of  its  parts  is  revocable  and  subject  to  further 
examination.  Therefore,  restored  forms  have  always  been  a  faithful 
reflection  of  the  general  conclusions  applicable  to  them.  The  Proto- 
Indo-European  word  for  'horse'  was  successively  posited  as  *akvas, 
*akivas,  *ekivos,  and  finally  *ekiWos;  only  a  and  the  number  of 
phonemes  have  remained  undisputed. 

The  aim  of  reconstruction  is,  then,  not  to  restore  a  form  for  its 
own  sake — this  would  be  rather  ridiculous  to  say  the  least — but  to 
crystallize  and  condense  a  set  of  conclusions  that  seem  logically  to 
follow  from  the  results  obtained  at  each  moment;  in  short,  its  aim 
is  to  record  the  progress  of  our  science.  No  one  has  to  defend  lin- 
guists against  the  rather  absurd  charge  of  intending  to  restore 
Proto-Indo-European  completely  as  if  they  wished  to  use  it.  They 
do  not  have  this  objective  even  in  studying  the  languages  that  are 
historically  attested  (one  does  not  study  Latin  linguistically  in 


order  to  speak  it  well).  There  is  even  less  justification  for  it  in  the 
case  of  individual  words  of  prehistoric  languages. 

Reconstruction,  though  always  subject  to  revision,  is  necessary 
for  an  overall  view  of  the  language  studied  and  of  its  linguistic  type. 
It  is  an  indispensable  instrument  for  depicting  with  relative  ease 
a  great  number  of  general  facts,  both  synchronic  and  diachronic. 
The  whole  set  of  reconstructions  immediately  illuminates  the  broad 
outlines  of  Proto-Indo-European.  For  instance,  we  know  that  suf- 
fixes were  formed  from  certain  elements  (t,  s,  r,  etc.)  to  the  exclu- 
sion of  others,  and  that  the  complicated  variety  of  the  vocahsm 
of  German  verbs  (cf.  werden,  wirst,  ward,  wurde,  worden)  obscures 
the  rules  governing  one  and  the  same  original  alternation :  e-o-zero. 
The  result  is  that  reconstruction  is  a  great  help  in  studying  the 
history  of  later  periods,  for  without  reconstruction  it  would  be 
much  more  difficult  to  explain  the  changes  that  have  occurred 
since  the  prehistoric  period. 

2.  Relative  Accuracy  of  Reconstructions 

We  are  absolutely  certain  of  some  reconstructed  forms,  but 
others  are  either  open  to  dispute  or  frankly  problematical.  We  have 
just  seen  that  the  accuracy  of  whole  forms  depends  on  the  relative 
accuracy  that  we  can  attribute  to  the  partial  restorations  that  go 
into  the  synthesis.  On  this  score  two  words  are  almost  never 
identical.  Between  Proto-Indo-European  forms  as  illuminating  as 
*esti  'he  is'  and  *diddti  'he  gives,'  there  is  a  difference,  for  the  re- 
duplicated vowel  of  the  second  form  gives  room  for  doubt  (cf. 
Sanskrit  dadati  and  Greek  didosi). 

There  is  a  general  tendency  to  consider  reconstructions  less 
accurate  than  they  actually  are.  Three  facts  should  fortify  our 

The  first  fact,  which  is  of  capital  importance,  was  mentioned 
earlier  (see  pp.  39  jEf.).  We  can  distinguish  clearly  the  sounds  of  a 
particular  word,  their  number,  and  their  delimitation.  We  have 
also  seen  (p.  54)  how  we  should  regard  the  objections  that  certain 
linguists  squinting  into  the  phonological  microscope  might  raise. 
In  a  sequence  Hke  ~sn-  there  are  doubtless  furtive  or  transitional 
sounds,  but  to  give  weight  to  them  is  antilinguistic ;  the  average 
ear  does  not  single  them  out,  and — even  more  important — speakers 


always  agree  on  the  number  of  elements  in  such  a  sequence.  We  can 
therefore  state  that  the  Proto-Indo-European  form  *ekiwos  had 
only  five  distinct,  differential  elements  to  which  speakers  had  to 
pay  heed. 

The  second  fact  has  to  do  with  the  system  of  the  phonological 
elements  of  each  language.  Any  language  operates  with  a  clearly 
delimited  gamut  of  phonemes  (see  p.  34).  The  least  frequent  ele- 
ments of  the  Proto-Indo-European  system  appear  in  no  fewer  than 
a  dozen  forms — and  the  most  frequent  in  a  thousand — all  attested 
through  reconstruction.  With  this  we  are  sure  of  knowing  them  all. 

Finally,  we  do  not  have  to  delineate  the  positive  qualities  of  the 
phonic  units  in  order  to  know  them.  We  must  consider  them  as 
differential  entities  that  are  characterized  by  their  being  distinct 
(see  p.  119).  This  is  so  basic  that  we  could  designate  the  phonic 
elements  of  an  idiom  that  is  to  be  reconstructed  by  numbers  or  by 
any  signs  whatsoever.  There  is  no  need  for  determining  the  abso- 
lute quality  of  e  in  *ekiwds  or  for  puzzling  over  whether  e  was  open 
or  closed,  just  how  far  forward  it  was  articulated,  etc.  All  this  is 
unimportant  unless  several  types  of  e  have  been  identified.  The 
important  thing  is  that  we  do  not  confuse  it  with  another  element 
singled  out  by  language  (a,  o,  e,  etc.).  This  is  another  way  of  saying 
that  the  first  phoneme  of  *ekiw6s  does  not  differ  from  the  second  of 
*medhyds,  the  third  of  *dge,  etc.,  and  that  without  specifying  its 
phonic  nature,  we  could  catalogue  it  and  assign  it  a  number  in  the 
table  of  Proto-Indo-European  phonemes.  The  reconstructed  form 
*ekiw6s  means  therefore  that  the  Proto-Indo-European  equivalent 
of  Latin  equos,  Sanskrit  agva-s,  etc.  was  composed  of  five  definite 
phonemes  taken  from  the  phonological  gamut  of  the  original  idiom. 

Within  the  limitations  just  outlmed,  reconstructions  do  retain 
their  full  value. 


Chapter  IV 


1.  Language  and  Race 

Thanks  to  his  retrospective  method,  the  Hnguist  can  go  back 
through  the  centuries  and  reconstruct  languages  that  were  spoken 
by  certain  nations  long  before  their  written  history  began.  But 
might  not  reconstructions  also  provide  information  about  the 
nations  themselves — their  race,  filiation,  social  relations,  customs, 
institutions,  etc.?  In  short,  does  language  provide  some  answers  to 
questions  that  arise  in  the  study  of  anthropology,  ethnography, 
and  prehistory?  Many  people  think  so,  but  I  believe  this  is  largely 
an  illusion.  Let  us  examine  briefly  some  parts  of  the  general 

First,  race.  It  would  be  wrong  to  assume  that  a  common  lan- 
guage implies  consanguinity,  that  a  family  of  languages  matches 
an  anthropological  family.  The  facts  are  not  so  simple.  There  is, 
for  instance,  a  Germanic  race  with  distinct  anthropological  charac- 
teristics: blond  hair,  elongated  cranium,  high  stature,  etc.;  the 
Scandinavian  is  its  most  perfect  example.  Still,  not  all  populations 
who  speak  Germanic  languages  fit  this  description;  thus  the  Ger- 
man from  the  foot  of  the  Alps  differs  strikingly  from  the  Scandi- 
navian. Might  we  at  least  assume,  however,  that  an  idiom  belongs 
exclusively  to  one  race,  and  that  if  nations  belonging  to  other  races 
use  the  idiom,  this  is  only  because  it  has  been  imposed  upon  them 
through  conquest?  No  doubt  nations  often  adopt  or  are  forced  to 
submit  to  the  language  of  their  conquerors  (e.g.  the  Gauls  after  the 
victory  of  the  Romans) ,  but  this  does  not  explain  everything.  For 
instance,  even  if  they  had  subjugated  so  many  different  popu- 
lations, the  Germanic  tribes  could  not  have  absorbed  all  of  them; 
we  would  have  to  imagine  a  long  period  of  preliistoric  domination 
and  still  other  unsubstantiated  circumstances. 


Consanguinity  and  linguistic  community  apparently  have  no 
necessary  connection,  and  we  cannot  draw  conclusions  from  one 
and  apply  them  to  the  other;  consequently,  in  the  numerous  in- 
stances where  anthropological  and  linguistic  evidence  do  not  agree, 
it  is  not  necessary  to  set  the  two  types  of  evidence  in  opposition  or 
to  choose  between  them ;  each  type  retains  its  own  value. 

2.  Ethnic  Unity 

What  can  we  learn  from  the  evidence  furnished  by  language? 
Racial  unity  alone,  a  secondary  force,  is  in  no  way  necessary  for 
linguistic  community.  But  there  is  another  type  of  unity — the  only 
crucial  type — which  is  of  infinitely  greater  importance  and  which 
is  constituted  by  the  social  bond:  ethnic  unity  [ethnisme].  By  this 
I  mean  a  unity  based  on  the  multiple  relations  of  religion,  civili- 
zation, common  defense,  etc.,  which  spring  up  even  among  nations 
of  different  races  and  in  the  absence  of  any  political  bond. 

Between  ethnic  unity  and  language  is  established  the  mutual 
relation  mentioned  earlier  (see  p.  20).  The  social  bond  tends  to 
create  linguistic  community  and  probably  imposes  certain  traits 
on  the  common  idiom;  conversely,  linguistic  community  is  to  some 
extent  responsible  for  ethnic  unity.  In  general,  ethnic  unity  always 
suffices  to  explain  linguistic  community.  For  example,  in  the  early 
Middle  Ages  a  Romance  ethnic  unity,  in  the  absence  of  any 
political  bond,  linked  nations  of  the  most  varied  origins.  Re- 
ciprocally, on  the  question  of  ethnic  unity,  we  must  first  consult 
language.  The  information  that  it  provides  takes  precedence  over 
everything  else.  Here  is  one  example.  In  ancient  Italy  the  Etrus- 
cans Hved  alongside  the  Latins.  If  we  try  to  determine  what  the 
two  nations  had  in  common  in  the  hope  of  tracing  them  back  to  the 
same  origin,  we  can  call  up  everything  that  they  transmitted 
(monuments,  reUgious  rites,  political  institutions,  etc.)  and  still 
lack  the  assurance  that  language  provides  immediately.  Four  lines 
of  Etruscan  are  enough  to  show  that  the  speakers  of  this  language 
belong  to  a  nation  distinct  from  the  ethnic  group  that  spoke  Latin. 

Thus  language — within  the  limitations  indicated — is  a  historical 
document.  That  the  Indo-European  languages  form  a  family,  for 
example,  is  proof  of  a  primitive  ethnic  unity  that  has  been  trans- 


mitted  more  or  less  directly  through  social  filiation  to  every  nation 
that  speaks  one  of  these  languages  today. 

3.  Linguistic  Paleontology 

Linguistic  unity  may  allow  us  to  predicate  social  community, 
but  does  language  reveal  the  nature  of  this  common  ethnic  unity? 

For  a  long  time  languages  were  considered  an  inexhaustible 
source  of  documents  concerning  the  nations  that  spoke  them  and 
their  prehistory.  Adolphe  Pictet,  a  pioneer  of  Celtism,  is  known 
especially  for  his  book  Les  Origines  indo-europeennes  (1859-63). 
His  work  has  served  as  a  model  for  many  others;  it  is  still  the  most 
engaging  of  all.  Pictet  looks  to  the  Indo-European  languages  for 
data  that  will  reveal  the  fundamental  traits  of  the  civilization  of 
the  "Aryans"  and  beUeves  that  he  can  fix  the  most  varied  details: 
material  things  (tools,  weapons,  domesticated  animals),  social  life 
(whether  they  were  a  nomadic  or  an  agricultural  nation) ,  family, 
government,  etc.  He  seeks  to  identify  the  cradle  of  the  Aryans, 
which  he  places  in  Bactriana,  and  studies  the  flora  and  fauna  of  the 
country  that  they  inhabited.  His  is  the  most  important  under- 
taking of  its  type.  The  science  that  he  founded  is  called  linguistic 

Other  efforts  in  the  same  direction  have  since  been  made.  One  of 
the  more  recent  is  Hermann  Hirt's  Die  Indogermanen  (1905-1907)  .^ 
Basing  his  research  on  the  theory  of  J.  Schmidt  (see  p.  209),  Hirt 
tries  to  identify  the  country  inhabited  by  the  Indo-Europeans.  But 
he  does  not  slight  linguistic  paleontology.  Lexical  facts  show  him 
that  the  Indo-Europeans  were  farmers,  and  he  refuses  to  place 
them  in  southern  Russia,  which  is  better  suited  to  nomadic  life.  The 
frequency  of  occurrence  of  names  of  trees,  especially  of  certain 
kinds  (fir,  birch,  beech,  oak),  makes  him  think  that  their  country 
was  wooded,  and  that  it  was  located  between  the  Harz  Mountains 
and  the  Vistula,  more  specifically  in  the  region  of  Brandenburg  and 
Berlin.  We  should  also  recall  that  even  before  Pictet,  Adalbert 

1  Cf.  also  d'Arbois  de  Jubainville,  Les  premiers  habitants  de  V Europe  (1877); 
O.  Schrader,  Sprachvergleichung  und  Urgeschichie  and  Reallexicon  der  indoger- 
manischen  A  Iterturnskunde  (works  that  appeared  a  little  earlier  than  the  volume 
by  Hirt);  and  S.  Feist,  Europa  im  lAchte  der  Vorgeschichte  (1910).  [Ed.] 


Kuhn  and  others  had  used  linguistics  to  reconstruct  the  mythology 
and  religion  of  the  Indo-Europeans. 

Now  we  cannot  expect  language  to  furnish  such  information  for 
the  following  reasons : 

First  is  the  uncertainty  of  etymology.  Scholars  have  at  last 
realized  how  rare  are  words  with  well-established  origins,  and  have 
become  more  cautious.  Here  is  an  example  of  the  rashness  that  once 
prevailed.  Given  servus  and  servare,  scholars  compared  the  two — 
they  probably  had  no  right  to  do  this — and  by  giving  the  first  word 
the  meaning  "guardian,"  they  were  able  to  conclude  that  a  slave 
was  originally  used  in  the  sense  of  "to  guard."  Nor  is  that  all.  The 
meanings  of  words  evolve.  The  meaning  of  a  word  often  changes 
whenever  a  tribe  changes  its  place  of  abode.  Scholars  were  also 
wrong  in  assuming  that  the  absence  of  a  word  proves  that  the 
primitive  society  knew  nothing  of  the  thing  that  the  word  names. 
Thus  the  word  for  "to  plow"  is  not  found  in  the  Asiatic  languages, 
but  this  does  not  mean  that  in  the  beginning  plowing  was  un- 
known; it  might  just  as  well  have  been  discarded  or  conducted  by 
other  procedures  known  by  different  names. 

The  possibility  of  loan-words  is  a  third  cause  of  uncertainty. 
An  object  that  is  borrowed  may  bring  its  name  along  with  it.  For 
instance,  hemp  came  into  the  Mediterranean  world  at  a  very  late 
date,  and  into  the  countries  to  the  north  even  later;  each  time,  the 
name  for  hemp  came  with  the  plant.  In  many  instances  the  absence 
of  extralinguistic  data  does  not  allow  us  to  ascertain  whether  the 
presence  of  the  same  word  in  several  languages  is  due  to  borrowing 
or  is  proof  of  a  common  original  tradition. 

The  foregoing  limitations  do  not  preclude  our  distinguishing 
with  no  hesitation  some  general  traits  and  even  certain  precise 
data.  For  example,  common  terms  indicating  kinship  are  abundant 
and  have  been  transmitted  very  clearly.  They  allow  us  to  state  that 
among  the  Indo-Europeans  the  family  was  a  complex  and  stable 
institution,  for  their  language  could  express  subtleties  that  ours 
cannot.  In  Homer,  eindteres  means  "sisters-in-law"  with  reference 
to  the  wives  of  several  brothers,  and  galooi  denotes  the  relation- 
ship between  the  wife  and  the  sister  of  the  husband.  Latin 
janitrlces  corresponds  to  eindteres  in  form  and  in  signification. 


Similarly,  "brother-in-law"  (the  husband  of  the  sister)  is  not 
named  by  the  same  word  as  "brothers-in-law"  (denoting  the  rela- 
tionship among  the  husbands  of  several  sisters).  Here  we  can 
identify  a  minute  detail,  but  usually  we  must  be  satisfied  with 
general  information.  The  same  applies  to  animals.  For  important 
species  like  the  bovine  we  can  rely  on  the  coincidence  of  Greek 
bous,  German  Kuh,  Sanskrit  gau-s,  etc.  and  reconstruct  the  Proto- 
Indo-European  form  *g20U-s;  besides,  the  inflection  of  the  word 
has  the  same  features  in  each  language,  and  this  would  be  impos- 
sible if  it  had  been  borrowed  from  another  language  at  a  later  date. 

Here  we  might  consider  another  morphological  fact  that  has  the 
dual  characteristic  of  being  limited  to  a  definite  zone  and  of  touch- 
ing upon  a  point  of  social  organization. 

In  spite  of  everything  that  has  been  said  about  the  relation  of 
dominus  and  domus,  linguists  do  not  seem  to  be  completely  satis- 
fied, for  the  use  of  the  suffix  -no-  in  forming  secondary  derivatives 
is  most  extraordinary.  There  are  no  formations  Uke  *oiko-no-s  or 
*oike-no-s  from  oikos  in  Greek,  or  *agva-na  from  agva-  in  Sanskrit. 
But  this  very  rarity  gives  the  suffix  of  dominus  its  value  and 
prominence.  Several  Germanic  words  are,  I  think,  quite  revealing : 

(1)  *\)eu'6a-na-z  'head  of  the  *\)eu'6d,  king,'  Gothic  \)iudans, 
Old  Saxon  ihiodan  {*\)eu'6d,  Gothic  \)iuda  —  Oscan  touto  'people'). 

(2)  *drux-ti-na-z  (partially  changed  to  *drux-ti-na-z)  'head  of 
the  *drux-ti-z,  army'  (whence  the  Christian  name  for  the  Master, 
i.e.  God),  cf.  Old  Norse  Drottinn,  Anglo-Saxon  Dryhten,  both  with 
final  -ina-z. 

(3)  *kindi-na-z  'head  of  the  *kindi-z  =  Latin  gens.^  Since  the 
head  of  the  gens  was  a  vice-ruler  with  respect  to  the  head  of  a 
*)}ewtSo,  the  Germanic  word  kindins  (completely  lost  elsewhere)  is 
used  by  ULfilas  to  name  the  Roman  governor  for,  in  his  Germanic 
way  of  thinking,  the  delegate  of  the  emperor  was  the  head  of  the 
clan  with  respect  to  the  Ipiudans;  however  interesting  the  associ- 
ation may  be  from  a  historical  viewpoint,  there  is  no  doubt  that 
the  word  kindins,  which  is  wholly  unlike  everything  Roman, 
indicates  a  division  of  the  Germanic  populations  into  kindi-z. 

Thus  the  secondary  suffix  -na-,  when  added  to  any  Proto- 
Germanic  theme,  means  'head  of  a  certain  community.'  All  that 
remains  now  is  to  observe  that  in  the  same  way  Latin  tribunus 


literally  means  'head  of  the  tribus,'  that  \>iudans  means  'head  of 
the  ]>iuda,'  and  finally,  that  dominus  means  'head  of  the  domus,' 
the  last  division  of  the  touta  =  piuda.  Dominus,  with  its  singular 
suffix,  seems  to  me  to  offer  almost  irrefutable  proof  not  only  of 
linguistic  community  but  also  of  a  community  of  institutions 
among  the  Italic  and  German  ethnic  groups. 

But  again  it  is  worth  noting  that  comparisons  between  languages 
rarely  yield  such  characteristic  indices. 

4.  Linguistic  Type  and  Mind  of  the  Social  Group 

Does  language,  even  if  it  fails  to  supply  much  precise  and 
authentic  information  about  the  institutions  of  speakers,  serve  at 
least  to  characterize  the  mind  of  the  social  group  that  speaks  it? 
A  popular  notion  is  that  a  language  reflects  the  psychology  of  a 
nation.  But  one  serious  objection  opposes  this  viewpoint:  psycho- 
logical causes  do  not  necessarily  underlie  linguistic  procedures. 

The  Semitic  languages  express  the  relation  of  a  substantival  de- 
terminant to  its  noun  (cf.  French  la  parole  de  Dieu  'the  word  of 
God')  by  simple  juxtaposition.  To  be  sure,  the  noun  that  is  de- 
termined has  a  special  form,  called  "construct  state,"  and  precedes 
the  determinant.  Take  Hebrew  dabar  'word'  and  ^elolilm}  'God': 
dabar  'elohim  means  'the  word  of  God.'  Should  we  say  that  such  a 
syntactical  pattern  reveals  something  about  the  Semitic  mind? 
That  would  be  a  rash  assertion,  for  Old  French  regularly  used  a 
similar  construction:  cf.  le  cor  Roland  'Roland's  horn,'  les  quatrc  fils 
Aymon  'Aymon's  four  sons,'  etc.  Now  the  procedure  arose  in 
•  Romance  through  sheer  chance,  morphological  as  well  as  phonetic : 
a  sharp  reduction  of  cases  forced  the  new  construction  on  the  lan- 
guage. It  is  entirely  possible  that  a  similar  accident  started  Proto- 
Semitic  on  the  same  route.  Thus  a  syntactical  fact  that  is  appar- 
ently one  of  its  indelible  traits  gives  no  accurate  clue  to  the  Semitic 

Another  example :  Proto-Indo-European  had  no  compounds  with 
a  word-initial  verbal  element.  That  German  has  such  compounds 
(cf.  Bethaus,  Springhrunnen,  etc.)  does  not  prove  that  at  a  given 
moment  the  Germans  modified  a  way  of  thinking  inherited  from 

2  The  symbol  [']  designates  the  alef  or  glottal  stop  that  corresponds  to  soft 
breathing  in  Greek.  [S.] 


their  ancestors.  We  have  seen  (p.  195)  that  the  innovation  was  due 
to  an  accident  which  was  not  only  material  but  also  negative,  the 
eUmination  of  the  a  in  betahus.  Everything  occurred  outside  the 
mind  and  in  the  realm  of  sound  changes,  which  readily  impose  a 
tight  yoke  on  thought  and  force  it  into  the  special  way  that  the  ma- 
terial state  of  signs  opens  to  it.  A  great  number  of  similar  observa- 
tions confirms  this  conclusion.  The  psychological  character  of  the 
linguistic  group  is  unimportant  by  comparison  with  the  elimination 
of  a  vowel,  a  change  of  accent,  or  many  other  similar  things  that 
may  at  any  moment  revolutionize  the  relation  between  the  sign 
and  the  idea  in  any  language  form  whatsoever. 

It  is  always  of  interest  to  determine  the  grammatical  character 
of  languages  (whether  historically  attested  or  reconstructed)  and 
to  classify  languages  according  to  the  procedures  that  they  use  for 
expressing  thought.  But  even  after  we  become  acquainted  with  the 
structures  of  languages  and  classify  them,  we  can  draw  no  accurate 
conclusions  outside  the  domain  of  linguistics  proper. 

Chapter  V 


We  have  just  seen  that  language  is  not  controlled  directly  by  the 
mind  of  speakers.  Let  me  emphasize,  in  concluding,  one  of  the 
consequences  of  this  principle:  no  family  of  languages  rightly 
belongs  once  and  for  all  to  a  particular  linguistic  type. 

To  ask  the  type  to  which  a  group  of  languages  belongs  is  to  for- 
get that  languages  evolve;  the  implication  is  that  there  is  an 
element  of  stability  in  evolution.  How  is  it  possible  to  impose 
limitations  on  an  activity  that  has  none? 

Of  course  many  people  really  have  in  mind  the  traits  of  the 
original  idiom  when  they  speak  of  the  characteristics  of  a  family; 

^  This  chapter,  though  it  does  not  deal  with  retrospective  linguistics,  is  in- 
cluded in  Part  Five  because  it  serves  as  a  conclusion  for  the  whole  work.  [Ed.] 


their  problem  is  not  insoluble  since  they  are  dealing  with  one 
language  and  one  period.  But  when  we  assume  that  there  are  per- 
manent traits  which  neither  time  nor  space  can  change  in  any  way, 
we  clash  head-on  with  the  fundamental  principles  of  evolutionary 
linguistics.  No  characteristic  has  a  right  to  permanent  existence; 
it  persists  only  through  sheer  luck. 

Take  the  Indo-European  family.  We  know  the  distinctive  traits 
of  the  language  from  which  it  derives.  The  sound  system  of  Proto- 
Indo-European  is  very  simple.  There  are  no  complicated  clusters 
of  consonants  or  double  consonants,  and  its  monotone  system 
gives  rise  to  an  interplay  of  extremely  regular  and  profoundly 
grammatical  alternations  (see  p.  157  and  p.  220) ;  the  tonic  accent 
can  in  principle  be  placed  on  any  syllable  in  a  word  and  therefore 
has  a  role  in  the  interplay  of  grammatical  oppositions ;  quantitative 
rhythm  is  based  solely  on  the  opposition  of  long  and  short  syllables; 
compounds  and  derivatives  are  easily  formed ;  nominal  and  verbal 
inflections  are  numerous;  and  the  inflected  word  with  its  self- 
contained  determiners  is  independent  in  the  sentence,  allowing 
much  freedom  of  construction  and  greatly  restricting  the  number 
of  grammatical  words  with  determinative  or  relational  value 
(preverbs,  prepositions,  etc.). 

Now  it  is  clear  that  none  of  the  foregoing  traits  has  been  retained 
in  its  original  form  in  the  different  Indo-European  languages,  and 
that  several  of  them  (e.g.  the  role  of  quantitative  rhythm  and  of 
tonic  accent)  no  longer  appear  m  any  member  of  the  group.  Some 
languages  have  even  changed  the  features  of  Proto-Indo-European 
to  such  an  extent  that  they  suggest  an  entu-ely  different  linguistic 
type  (e.g.  English,  Armenian,  Irish,  etc.). 

It  would  be  more  fitting  to  speak  of  certain  transformations  that 
affect  different  languages  belonging  to  the  same  family.  For  in- 
stance, progressive  weakening  of  the  inflectional  mechanism  is 
characteristic  of  the  Indo-European  languages,  although  they  all 
offer  striking  differences.  Slavic  has  put  up  the  strongest  resistance 
while  EngUsh  has  reduced  inflection  almost  to  zero.  To  offset  this,  a 
rather  stable  word-order  has  developed,  analytical  processes  of 
expression  have  tended  to  replace  synthetic  processes,  prepositions 
express  case  values  (see  p.  180),  auxiliaries  have  taken  the  place  of 
compound  verbal  forms,  etc. 


We  have  seen  that  a  trait  of  the  prototype  may  not  appear  in 
some  of  the  derived  languages.  The  reverse  is  equally  true.  It  is  not 
unusual  even  to  find  that  the  common  traits  of  all  the  representa- 
tives of  a  family  do  not  appear  in  the  original  idiom.  This  is  true  of 
vocalic  harmony  (i.e.  similarity  of  some  type  between  the  timbre 
of  every  suffixed  vowel  and  the  last  vowel  of  the  radical).  This 
salient  trait  is  found  in  Ural-Altaic  (a  large  group  of  languages 
spoken  in  Europe  and  Asia  and  extending  from  Finland  to  Man- 
churia) but  is  probably  due  to  later  developments.  Vocalic  har- 
mony is  then  a  common  trait  but  not  an  original  one ;  consequently 
we  cannot  invoke  it — any  more  than  agglutination — to  prove  the 
common  origin  (highly  debatable)  of  these  languages.  We  also 
know  that  Chinese  has  not  always  been  monosyllabic. 

The  thing  that  first  strikes  us,  when  we  compare  the  Semitic 
languages  with  their  reconstructed  prototype,  is  the  persistence  of 
certain  traits.  The  Semitic  languages  seem,  more  than  any  other 
family,  to  constitute  a  type,  unchangeable  and  permanent,  with 
traits  of  the  family  inherent  in  each  language.  The  following  traits, 
many  of  which  contrast  sharply  with  those  of  Proto-Indo-Euro- 
pean, set  Proto-Semitic  apart.  Compounds  are  practically  non- 
existent. Derivation  plays  only  a  small  part.  The  mflectional 
system  is  poorly  developed  (better  in  Proto-Semitic,  however,  than 
in  the  daughter  languages) ,  with  the  result  that  strict  rules  govern 
word-order.  The  most  notable  trait  has  to  do  with  the  structure  of 
the  root  (see  p.  187).  It  regularly  includes  three  consonants  (e.g. 
q-t-l  'kill')  which  are  retained  in  every  form  within  a  given  language 
(cf.  Hebrew  qaial,  qdtld,  qtol,  qitll,  etc.),  and  which  do  not  change 
from  one  language  to  another  (cf.  Arabic  qatala,  qutila,  etc.).  In 
other  words,  consonants  express  the  "concrete  sense"  or  lexical 
value  of  words  while  vowels — with  the  help  of  certain  prefixes  and 
suffixes,  of  course — have  the  exclusive  role  of  indicating  gram- 
matical values  through  the  interplay  of  their  alternations  (e.g. 
Hebrew  qatal  'he  killed,'  qtol  'to  kill';  with  a  suffix,  qtdl-u  'they 
killed' ;  with  a  prefix,  ji-qtol  'he  will  kill' ;  and  with  both,  ji-qtl-u 
'they  will  kill,'  etc.). 

Against  the  foregomg  facts,  and  in  spite  of  the  statements  that 
they  have  elicited,  we  must  defend  our  principle :  there  are  no  un- 


changeable  characteristics.  Permanence  results  from  sheer  luck; 
any  characteristic  that  is  preserved  in  time  may  also  disappear  with 
time.  But  to  come  back  to  Semitic.  We  see  that  the  "law"  of  the 
three  consonants  is  not  really  characteristic  of  the  Semitic  family 
since  analogous  phenomena  appear  in  other  families.  In  Proto- 
Indo-European,  rigid  laws  also  govern  the  consonantal  structure 
of  roots.  For  example,  two  sounds  of  the  series  i,  u,  r,  I,  m,  n  never 
follow  e;  a  root  like  *serl  is  impossible.  The  functioning  of  Semitic 
vowels  is  even  more  instructive.  Indo-European  has  an  equally 
rigid  but  less  rich  set  of  vowels;  oppositions  like  Hebrew  dabar 
'word,'  dbdr-im  'words,'  dibre-hem  'their  words,'  etc.  recall  German 
Gast:  Gdste,  fiiessen:  floss,  etc.  In  both  instances  the  genesis  of  the 
grammatical  procedure  is  the  same.  Mere  phonetic  modifications, 
which  are  due  to  blind  evolution,  result  in  alternations.  The  mind 
seizes  upon  the  alternations,  attaches  grammatical  values  to  them, 
and  spreads  them,  using  the  analogical  models  which  chance 
phonetic  developments  provide.  The  immutability  of  the  three  con- 
sonants in  Semitic  is  only  a  general  rule,  not  a  hard-and-fast  one. 
We  could  be  sure  of  this  a  priori,  but  our  view  is  confirmed  by  the 
facts.  In  Hebrew,  for  example,  the  root  ^ands-im  'men'  has  the 
three  expected  consonants,  but  its  singular  'is  has  only  two,  for 
this  is  the  reduced  form  of  the  older  form  that  contained  three 
consonants.  Even  if  we  agree  that  Semitic  roots  are  quasi-immuta- 
ble, this  does  not  mean  that  they  have  an  inherent  characteristic. 
It  means  simply  that  the  Semitic  languages  have  suffered  fewer 
phonetic  alterations  than  many  others,  and  that  consonants  have 
been  better  preserved  in  this  group  of  languages  than  elsewhere.  We 
are  dealing  with  something  evolutionary  and  phonetic,  not  some- 
thing grammatical  or  permanent.  To  proclaim  the  immutability  of 
roots  is  to  say  that  they  have  undergone  no  phonetic  change, 
nothing  more,  and  we  cannot  vow  that  changes  will  never  occur. 
Generally  speaking,  everything  that  time  has  done,  time  can  undo 
or  change. 

We  now  realize  that  Schleicher  was  wrong  in  looking  upon  lan- 
guage as  an  organic  thing  with  its  own  law  of  evolution,  but  we 
continue,  without  suspecting  it,  to  try  to  make  language  organic 
in  another  sense  by  assuming  that  the  "genius"  of  a  race  or  ethnic 


group  tends  constantly  to  lead  language  along  certain  fixed  routes. 
From  the  incursions  we  have  made  into  the  borderlands  of  our 
science,  one  lesson  stands  out.  It  is  wholly  negative,  but  is  all  the 
more  interesting  because  it  agrees  with  the  fundamental  idea  of 
this  course:  the  true  and  unique  object  of  linguistics  is  language 
studied  in  and  for  itself. 



Ablaut,  158  f.;  160 

Accent,  86,  58 

Agglutination,  definition  of  — ,  176; 
three  phases  in  — ,  177;  —  con- 
trasted with  analogy,  177  f.;  — 
always  precedes  analogy,  179  note 

Alphabet,  see  Writing;  borrowed  — , 
28;  Greek  — ,  27,  39 

Alteration  of  the  sign,  74  f .,  linguis- 
tic —  is  always  partial,  84,  87 

Alteration,  157 ;  definition,  158  f . ; 
laws  of  — ,  158;  —  is  synchronic 
and  grammatical,  159;  —  tightens 
the  grammatical  bond,  160  f . 

Analogy,  161-173;  its  importance, 
171 ;  —  counteracts  phonetic 
changes,  161 ;  —  as  a  proportion, 
161,  165,  167;  —  and  formative 
elements,  162,  170;  mistakes  con- 
cerning — ,  162  f.;  —  is  creation 
and  not  change,  163  f . ;  its  mecha- 
nism, 163 ;  —  is  grammatical,  165 ; 

—  originates  in  speaking,  165,  168 ; 
two  theories  of  — ,  166  f.;  —  as 
a  force  in  evolution,  169,   171  f.; 

—  indicates  changes  in  interpreta- 
tion, 169  f . ;  —  as  a  conservative 
force,  172  f.;  —  contrasted  with 
folk  etymology,  173  f.;  —  con- 
trasted  with   agglutination,    177  f . 

Analysis,  objective  — ,  183  f.;  sub- 
jective — ,  183  f.;  —  and  the  de- 
fining of  subunits,  185  f . 

Anthropology  and  linguistics,  6,  222. 

Aperture,  basis  for  classifying 
sounds,  44  f. ;  —  and  opening  and 
closing  sounds,  52 

Aphasia,  10  f . 

Apparatus,  vocal  — ,  41  f . 

Arbitrary  nature  of  the  sign,  defini- 
tion, 67  f.;  —  and  the  immutabil- 
ity of  language,  72  f.;  —  and  al- 
teration, 74  f . ;  absolute  and  rela- 
tive  arbitrariness,    131  f.;   —  and 

phonetic  change,  151,  161;  —  and 
analogy,  166 

Articulation  and  auditory  impres- 
sion, 8;  two  meanings  of  — ,  10, 
113;  diversity  of  oral  — ,  42  f.; 
value  of  oral  —  in  classifying 
sounds,  44;  sistants  or  holds,  52 
and  note;  muscular  image  of  — , 

Aspects  of  the  verb,  117 

Associative  faculty,  13 

Atlas,  linguistic  — ,  202 

Aufrecht,  3 

Beats,     homogeneous     —     of     the 

spoken  chain,  38 
Benfey,  3 
Berthelot,  25 

Bopp,  2,  3,  24,  82,  184,  215 
Boundary,  syllabic  — ,  57 
Braune,  5 
Broca,  10 
Brugmann,  5 

Cartography,  linguistic  — ,  202 

Cavity,  oral  — ,  41 ;  nasal  — ,  42. 

Chain,  phonic  (or  spoken)  — ,  38  f.; 
49  f.;  51  f. 

Changes  in  language,  originate  in 
speaking,  18,  98;  —  are  always 
partial,  84  f.;  87 

Changes,  morphological  — ,  93; 
semantic  — ,  93;  syntactical  —  93 

Changes,  phonetic  — ,  143-161 ;  — 
are  unrelated  to  the  system  of  lan- 
guage, 18  f.;  —  affect  sounds,  not 
words,  93;  their  regularity,  143; 
absolute  and  conditioned,  spon- 
taneous and  combinatory  — , 
144  f.;  see  Phonetics 

Chess,  compared  with  the  system  of 
language,  22  f .,  88,  110 

Children,  role  of  —  in  phonetic 
evolution,  149 




Climate     and     linguistic     changes, 

147  f.,  198 
Community  of  speakers,  78 
Comparison  of  unrelated  languages, 

192   — ;    of   related   languages,  4, 

199,  218 
Comparative  philology,  mistakes  of 

the  school  of  —  3  f .,  24  f .,  163,  184, 

209  f.,  215 
Compounds,    products    of    analogy, 

178  and  note;  Germanic  — ,  141  f., 

227  f . ;     Proto-Indo-European    — , 

178  note,  227  f . 
Concept,  12,  66;  called  signified,  67, 

103,  113  f. 
Consanguinity   and   linguistic   com- 
munity, 222 
Consonants,  48,  57  f.;  middle  —  or 

tenues,  35 
Construction  and  structure,  178 
Co-ordinating  faculty,  13 
Cords,  vocal  — ,  41 
Curtius,  3 

Darmsteter,  32 

Delimiting  of  linguistic  units,  104  f . ; 
—  of  phonemes,  38 

Dentals,  45 

Derivatives,  products  of  analogy, 

Deschamps,  25 

Diachrony,  81;  see  also  Linguistics, 

Dialectal,  borrowed  —  forms,  156 

Dialects,  natural  —  do  not  exist, 
202;  distinction  between  —  and 
languages,  203  f.;  —  and  literary 
language,  21,  195 

Diez,  5 

Differences,  role  of  —  in  creating 
values,  114  f.,  117  f.;  there  are 
only  —  in  language,  120 

Differentiation,  linguistic  —  on  con- 
tinuous territory,  199  f . ;  —  on 
separate  territories,  208  f . 

Diphthong,  implosive  link,  61;  "as- 
cending" — ,  61 

Diversity  of  languages,  191  f . ;  — 
among  related  languages,  191, 
197;  absolute  — ,  192 

Dominus,  etymology  of  — ,  226 

Doublets,  nonphonetic  character  of 

— ,  115  f. 
Dualities,  linguistic  — ,  8 

Economy,  political  — ,  79 

Entities,  concrete  —  of  language, 
102  f . ;  abstract  — ,  137  f . 

Ethnic  unity,  223  f.;  Italic  and  Ger- 
man — ,  226 

Ethnography  and  linguistics,  6,  20, 

Etruscans  and  Latins,  223 

Etymology,  folk  — ,  173  f.;  —  with 
and  without  deformation,  174;  in- 
complete — ,  174  f.;  —  compared 
with  analogy,  174  f . 

Etymology  and  orthography,  28,  31 ; 
uncertainty  of  — ,  225  f.;  defini- 
tion, 173 

Evolution,  linguistic  — ,  8 ;  —  begins 
in  speaking,  18,  98;  —  of  gram- 
matical facts,  142;  see  Changes, 

Expiration,  42  f . 

Explosion,  51  f.;  duration  of  — ,  60 

Extension,  geographical  —  of  lan- 
guages, 21;  see  Linguistics,  geo- 

Facts,  grammatical  —  and  linguistic 
units,  122 

Faculty  of  speech,  9f. 

Families  of  languages,  6,  191  f.; 
Indo-European  family,  2,  204  f ., 
209  f.;  Bantu  — ,  192;  Finno-Ugric 
— ,  192;  —  have  no  permanent 
traits,  228  f.;    Ural-Atlaic  — ,  230 

Fashion,  75,  151 

Formulas,  articulatory  —  for 
sounds,  44 

Fortuitous  character  of  a  language- 
state,  85 

Fricatives,  46 

Furtive  sounds,  54  f .,  220 

Gillieron,  31  note,  32  note,  202 
Glottis,  41  f. 
Gothic,  216 

Grades  of  the  vocalic  system,  4 
Grammar,     definition,     134;     tradi- 
tional or  classical  — ,  1,  82;  —  is 



normative  and  static,  1,  82;  gen- 
eral — ,  100;  "historical"  — ,  134, 
142,  143  note 

Grammont,  32  note 

Grimm,  3,  25 

Gutturals,  44,  46;  palatals,  44,  46; 
velars,  45,  46  and  note 

h,  aspirate  — ,  48 ;  in  French,  32  f . 

Harmony,  vocalic  —  of  the  Ural- 
Altaic  languages,  230 

Hiatus,  59  f. 

Hirt,  224 

History  of  linguistics,  If.,  81  f.;  re- 
lations between  political  —  and 
phonetic  changes,  150 

Holds  or  sistants,  52  and  note 

Identity,  synchronic  — ,  107  f . ;  dia- 

chronic  — ,  181  f . 
Idiom,   191  f. 

Immutability   of  the  sign,  71  f . 
Implosion,  51  f.;  duration  of  — ,  60  f. 
Indirect  spellings,  29;  fluctuating  — , 

29  f . ;  see  Writing 
Indo-European,     characteristics     of 

— ,  229 
Inflection,  185;  zero  — ,  185 
Institution,  language  is  a  social  — , 

10,  16 
Intercourse  or  unifying  force,  205  f.; 

two  forms  of  — ,  206 
Interjections,  69 
Isogloss  lines,  203 

Jespersen,  40  note,  42  note 
Jones,  2 

Koine  or  literary  Greek,  196 
Kuhn,  3,  204,  224 

I,  dental,  guttural,  nasal,  and  pa- 
latal —  47 

Labials,  45 

Labio-dentals,  46 

Language,  norm  of  the  facts  of 
speech,  9;  —  is  social,  homogen- 
ous, and  concrete,  14  f.;  —  is  dis- 
tinct from  speaking,  14  f .,  17  f .,  77, 
165;  —  is  not  a  name-giving  sys- 
tem, 16,  65;  —  and  speaking  are 

interdependent,  18;  how  —  exists, 
19;  —  is  a  form,  not  a  substance, 
113,  122;  languages  and  dialects, 

Languages,  Germanic  — ,  5,  216; 
Romance  — ,  5,  213,  217;  Semitic, 
227,  230 

Larch,  96 

Larynx,  41  f. 

Lateral  consonants,  47  f. 

Lautverschiebung ,  see  Mutation, 

Law,  Verner's  — ,  145 

Laws,  linguistic  — ,  91  f . ;  synchronic 
—  are  general  but  not  imperative, 
92  f.;  diachronic  —  are  impera- 
tive but  not  general,  93;  phonetic 
— ,  93  f . ;  wrong  statement  of 
phonetic  — ,  145  f.;  —  of  alterna- 
tion, 158 

Least  effort,  cause  of  phonetic 
changes,  148 

Leskien,  5 

Lexicology,  a  part  of  grammar,  135 

Limiting  of  arbitrariness,  basis  for 
the  study  of  language,  133  f . 

Linguistics  is  a  part  of  semiology, 
15  f. ;  —  of  language  and  —  of 
speaking,  see  Language;  external 
and  internal  — ,  20  f.;  synchronic 
or  static  — ,  81,  99  f.;  "historical," 
diachronic,  or  evolutionary,  81, 
99,   140  f.;    geographical  — ,   191  f. 

Liquids,  44,  47  f. 

Lithuanian,  24,  216 

Loan-words,  21,  36,  155  f.,  225 

Mechanism  of  language,  127  f.,  130, 

Meillet,  92  note 

Meter,  36 

Method,  comparative  — ,  3f.;  —  of 
external  and  of  internal  Unguis- 
tics,  22  f.;  —  of  synchronic  and 
diachronic  linguistics,  90  f.;  pros- 
pective and  retrospective  — ,  212  f . 

Migrations,  204  f.;  theory  of  — ,  209 

Millardet,  202  note 

Morphology,  inseparable  from  syn- 
tax, 135 

Morris,  16  note 



Motivation,  131  f. 

Movements,  facilitating  articulatory 

— ,  55 
Miiller,  3 

Mutability  of  the  sign,  74  f. 
Mutation,   consonantal  — ,  25,   144, 


Names  denoting  kinship   in  Proto- 
Indo-European,  225  f . 
Nasalized  sounds,  43 
Nasals,  45;  voiceless  — ,  45 
Naville,  16  note 
Neogrammarians,  5,  184 
Non-sonants,  57  f. 
Nyrop,  36 

Occlusives,  45  f . 

Old,    three    meanings   of   the   word 

—  applied  to  language,  215  f. 
Old   Slavic,  22,  217 
Onomatopoeia,  69 
Opposition  and  difference,  121 
Orthography,  25  f . ;  see  Writing  and 

Osthoff,  5 
Opening  sounds,  52 

Palatals,  46  f. 

Palate,  41 

Paleontology,  linguistic  — ,  224 

Panchronic  viewpoint,  95  f . 

Paradigms,  inflectional  — ,  127 

Participle,  present  — ,  96 

Parts  of  speech,  109,  138 

Paul,  5 

Peak,  vocalic  — ,  57 

Permutation,  synonym  of  alterna- 
tion, 160 

Perspective,  synchronic  and  dia- 
chronic  — ,81,  87,  90;  prospective 
and  retrospective  — ,  212  f. 

Philology,  method  of  — ,  1,  7;  com- 
parative — ,  2 

Phonation,  unrelated  to  language,  18 

Phonemes,  fixed  number  of  — ,  15, 
34,  40,  119,  220;  their  delimita- 
tion, 38,  42  f.;  their  description, 
39  f.;  —  are  differential,  54,  119, 
221 ;  —  and  sounds,  66 ;  their  syn- 

tagmatic  and  associative  relations, 
130  f. 

Phonetics,  32  f.;  —  and  grammar, 
17  f .,  152 ;  phonetic  means  non- 
significant, 18,  140  f.;  —  is  a  part 
of  diachronic  linguistics,  140 

Phonographic   recordings,  23 

Phonological  species,  40,  53 

Phonology,  32-64;  wrongly  called 
phonetics,  32  f. ;  —  is  a  part  of 
speaking,  33;  combinatory  — , 
50  f. 

Physiology  and  linguistics,  7 

Physiology  of  sounds,  see  Phonol- 

Pictet,  216,  224 

Plural  and  dual,  116 

Polite  formulas,  68 

Pott,  3 

Prefix,  187  f. 

Prehistory  and   linguistics,  6,  223  f. 

Prepositions,  unknown  in  Proto- 
Indo-European,   180 

Preservation  of  linguistic  forms,  173 

Preverbs,  unknown  in  Proto-Indo- 
European,   180 

Procedure  and  process,  distinction 
between  — ,  176 

Pronunciation  and  writing,  29  f. ;  — 
determined  by  etymology,  31 ;  — 
corrupted  by  writing,  31  f.;  rela- 
tive freedom  of  — ,  119 

Proto-Indo-European,  228  f . 

Provincialism  and  intercourse,  205  f . 

Psychology,  social  —  and  linguistics, 
7,  16 

r,  trilled  —  and  burr,  47 

Race  and  language,  222  f. 

Radical  or  theme,  185  f . 

Reading  and  writing,  34 

Reality,  synchronic  — ,  109;  dia- 
chronic — ,  181 

Reconstruction,   linguistic  — ,  218  f. 

Relations,  sjmtagmatic  and  associa- 
tive — ,  122  f . ;  their  interdepend- 
ence, 128  f.;  their  role  in  determ- 
ining phonemes,  130  f.;  —  are 
the  basis  for  the  divisions  of 
grammar,   136  f . 

Rhotocization,  144,  146 



Ritschl,  1 

Root,  definition,  186;  characteristics 
of  the  —  in  German,  186  f . ;  —  in 
French,  187;  —  in  the  Semitic 
languages,  187,  230 

Roudet,  40  note 

Sanskrit,  discovery  of  — ,  2  f.;  exag- 
gerated importance  of  — ,  215  f.; 
age  of  — ,  216 

Schleicher,  3,  4,  231 

Schmidt,  203,  209,  224 

Semantics,  16  note 

Semiology,  definition,  16;  based 
mainly  on  systems  of  arbitrary 
signs,  67  f . 

Semi-vowels,  48 

Sentence,  124;  —  as  a  unit,  106;  — 
equivalents,  128 

Separation,  geographical  —  and  lin- 
guistic  differentiation,   208  f . 

Shift  in  the  relationship  between 
the  signifier  and  the  signified,  75 

Sievers,  5,  40  note,  58,  61,  62 

Sign,  linguistic  — ,  its  composition, 
66  f.;  its  immutability,  71  f.;  its 
mutability,  74  f.;  —  considered  in 
its  totality,  120  f.;  unmotivated 
and  relatively  motivated  signs, 
131  f.;  zero  — ,  87,  118,  185,  187 

Signification,  114  f. 

Signified,  67,  102  f.;  see  Signifier 

Signifier,  definition,  67 ;  linearity  of 
— ,  70,  123 ;  —  exists  only  through 
the  signified  and  vice  versa.  102  f . 

Silbenbildend  and  silbisch,  59,  61 

Sociology  and  linguistics,  6 

Solidarities,  syntagmatic  and  associ- 
ative — ,  127  f .,  133 

Sonant,  57 ;  Proto-Indo-European 
sonants,  51,  63 

Sonority  of  phonemes,  43;  role  of 
—  in  syllabication,  58 

Sound,  complexity  of  — ,8;  —  and 
auditory  impression,  38  f.;  laryn- 
geal — ,  42  f.;  —  and  noise,  48;  — 
is  not  language,  110 

Sound-image,  12,  15,  66  note;  —  is 
psychological,  66;  —  is  called  sig- 
nifier, 67 

Sounds,    classification    of    — ,    43  f.; 

closing  and  opening  — ,  52;  fur- 
tive — ,  54  f .,  220 ;  shapelessness  of 
— ,  111. 

Speaking,  an  individual  act,  14;  — 
is  distinct  from  language,  see  Lan- 
guage; mode  of  existence  of  — , 
19;  —  is  the  seat  of  linguistic 
change,  19,  98,  143  note,  168 

Speaking-circuit,  11  f. 

Species,  phonological  — ,40;  —  are 
abstract,  53  f . 

Speech,  language  and  speaking,  77; 
—  is  heterogenous,  9;  —  is  a 
natural  faculty,  9f.;  —  is  articu- 
lated, 10 

Speech,  parts  of  — ,  109,  138 

SpeUing,  indirect  — ,  29;  fluctuating 
— ,  29  f.;  see  Writing 

Spirants,  46  f . 

Stability,  political  —  and  phonetic 
changes,   150  f. 

Substratum,  linguistic  —  and  phon- 
etic  changes,   151 

Subunits  of  words,  106,  127  f.,  129, 
185  f. 

Suffix,  178;  zero  — ,  186 

Syllable,  50,  57  f. 

Syllabic  boundary,  57  f . 

Symbol,  contrasted  with  sign,  68  f . 

Synchrony,  see  Linguistics,  syn- 

Syntagm,  definition,  122  f.;  see  Re- 
lations, syntagmatic 

Syntax,   135  f. 

System  of  language,  8,  22  f.,  72  f., 
79,  113,  133;  see  Mechanism. 

System,  phonological  — ,  34  f.,  221. 

Tense,  116  f. 

Terminology,  linguistic  —  is  in- 
exact, 5  note;  phonological  —  is 
imperfect,  44 

Theme  or  radical,  185  f. 

Thought,  shapelessness  of  — ,  111 

Time,  effect  of  —  on  language,  74, 
78,  197  f. 

Trombetti,   192 

Type,  linguistic  —  and  mind  of  the 
social  group,  227  f.;  —  and  fami- 
lies of  languages,  228  f. 



Umlaut,  24,  83  f .,  157 

Units,  linguistic  — ,  103  f . ;  complex 
— ,  105  f .,  124 ;  problems  in  defin- 
ing — ,  llOf.;  importance  of  — , 
110  f.;  differential  character  of  — , 
121  £.;  —  and  grammatical  facts, 
122,  179;   diachronic  — ,  181 

Unsilbisch,  61 

Uvula,  41  f. 

Value,  linguistic  — ,  110  f.;  its  con- 
ceptual side,  114  f.;  —  is  distinct 
from  signification,  114;  its  ma- 
terial side,  117  f. 

Velars,  46 

Verner's  law,  145  f. 

Versification,  36 

Vibrants,  47 

Vibrations,  lar5mgeal  — ,  42  f . 

Vocalic  peak,  57 

Vowels  contrasted  with  consonants, 
48;  —  contrasted  with  sonants 
57  f.;  open  and  closed,  whispered, 
and  voiceless  — ,  48  f . 

Waves,  innovating  — ,  203,  206 
Weigand,  202  note 

Wenker,  202 

Wellentheorie,  209 

Whitney,  5,  10,  76 

Wolf,   1 

Words  and  units  contrasted,  105  f ., 
113  f. 

Word-unit  and  phonetic  changes,  94 

Writing  and  language,  15;  necessity 
for  studying  — ,23;  —  is  distinct 
from  language,  23;  —  is  not  nec- 
essary for  linguistic  stability,  24; 

—  and  the  literary  language,  25; 

—  changes  less  frequently  than 
language,  27;  etymological  — ,  28; 
troubles  caused  by  — ,  29;  phono- 
logical — ,  33  f .  interpretation  of 
— ,  34  f . ;  recording  of  implosion 
and  explosion,  52  f.,  60  f.;  system 
of  —  compared  with  the  system 
of  language,  119  f. 

Writing,  systems  of  —  25  f.;  ideo- 
graphic —  (Chinese)  and  pho- 
netic — ,  26;  syllabic  —  (Cypri- 
ots),  26,  39,  50;  consonantal  — 
(Semites),  39 

Zend,  22 

Zones,  dialectal,  199  f. 

University  of  California  Library 
Los  Angeles 

This  book  is  DUE  on  the  last  date  stamped  below. 


MAR  1  7  2005 



.mN  1  8 



^  NOV  26 '91 

L  005  883  998  6